Just Ask Replit: The Future of Knowledge Work (Live Deep Dive)

Video authored by the channel shown, not by IANIA; all rights belong to the creator. Transcript and summary generated automatically; the original-language content is preserved.

Original language: English

Original video: Just Ask Replit: The Future of Knowledge Work (Live Deep Dive) · Channel: Replit · 1:54:09 · Watch on YouTube

Topics: AI agents, Task and workflow automation, Data, databases and embeddings, Software engineering with AI

Summary

Replit hosts a live deep dive on how AI is changing knowledge work, arguing that people should use AI to amplify their judgment and expertise rather than simply automate isolated tasks. The presenters demonstrate moving from broad prompts to recurring routines that combine calendar, Slack, meeting notes, and other connected sources, then refining the outputs through human review. They also analyze Aura’s financial data with interactive charts, web research, and an investor briefing, showing how an exploratory chat can lead to more durable tools or reports. Throughout, they emphasize that workers remain responsible for accuracy, quality, and the recommendations they share.

Key points

Knowledge work turns information into decisions, ideas, and solutions; AI can accelerate execution, but human judgment and quality control remain essential.
Start with the underlying need behind a task rather than merely reproducing its current output—for example, consider whether real-time alerts could serve better than a weekly report.
Think in systems and iterate: test a prompt, assess the result, refine the context, and automate or build a more durable solution only when its value is clear.
Replit routines can run on a schedule in the cloud, without the user being logged in, and can use connected sources such as Slack, calendars, Granola, and Notion.
Integrations and permissions determine what context the agent can access; the presenters recommend starting with connectors and narrowing access and prompts as needed.
A practical example shows a manual live-chat collection process evolving into an automated tool that gathers, analyzes, and summarizes stream data.
The agent can analyze uploaded financial data, generate interactive charts, research comparisons, and produce reports or presentations from its findings.
AI outputs need to be checked and edited: users are responsible for validating sources, data, and recommendations before sharing or acting on them.
The presenters argue that Replit supports a path from prototype to production, including deployment and infrastructure, rather than leaving users with a local-only prototype.
They describe the emerging full-stack knowledge worker as someone able to take an idea from initial insight through building and delivering something useful.

Chapters

Start	Chapter	Summary
00:01	Introduction and show plan	The hosts introduce the live discussion on knowledge work, AI prompting, and Replit demonstrations.
06:54	What knowledge work means in the AI era	Rhyar defines knowledge work and argues that AI amplifies both human expertise and the quality—or poor quality—of what people produce.
17:20	How AI changes roles and tasks	The hosts discuss how automation can free workers to take on broader, more strategic responsibilities rather than simply eliminating jobs.
25:41	From single prompts to recurring briefings	Rhyar compares basic, intermediate, and advanced prompts, then asks Replit to create a recurring briefing using connected work sources.
37:34	Testing and refining a daily routine	The hosts review the initial briefing, discuss how to tune its sources, and explain that routines can run in the cloud without a deployment.
46:23	From manual chat collection to an automated dashboard	Rhyar traces how manually copying livestream comments became an automated system that organizes feedback, transcripts, and show insights.
52:35	Analyzing financial data with charts	The hosts upload Aura financial data and prompt the agent to chart revenue and paid-member trends before expanding the analysis.
1:00:21	Exploring chart types and market comparisons	They discuss chart customization and use the agent to investigate financial performance and potential market benchmarks.
1:08:04	Research, judgment, and AI quality control	The hosts emphasize using AI for research while checking its evidence, refining the analysis, and taking responsibility for the final recommendation.
1:17:19	Investigating Aura’s delayed IPO	They ask the agent to research why Aura delayed its IPO and examine the resulting analysis as a starting point rather than a definitive answer.
1:23:29	From prototype to production	Rhyar shares a story about a nontechnical friend’s local prototype to illustrate the challenges of deployment, access control, and scaling.
1:35:32	Creating an investor briefing	The agent turns the analysis into an investor presentation, and the hosts review its financial summaries and caveat that it is not financial advice.
1:42:14	Reviewing outputs and validating analysis	The hosts discuss how to present findings, verify data and sources, and review the generated briefing and charts.
1:47:22	Closing thoughts on full-stack knowledge work	The hosts recap the demonstrations and describe how AI can help knowledge workers automate tasks, move up the value chain, and carry ideas into practice.
1:53:05	Sign-off	The hosts thank viewers, mention upcoming community events, and close the livestream.

Quotes

AI is accelerating execution, but you still need to bring that taste and enforce the quality — 09:30

Now you have the ability to kind of build a real-time feedback loop — 12:07

within 20 minutes of the show ending, it’s done. — 23:34

you are still responsible. You’re like the editor now. — 1:14:14

The full stack knowledge worker now is the person that has a little bit of the engineering knowledge — 1:51:03

Concepts and entities

Replit, Rhymar, Manny, knowledge work, first-principles thinking, routines, Granola, Slack, Notion, MCP, ADA, Sankey chart, Reream, Aura, Whoop

Full transcript (original language)

00:01 Let it [music] feel. [music] You meant [music] [music] [music]

00:38 [music] Let it all feel. All right. What’s up everybody? Welcome

01:09 back to another episode of Replet Live. My name is Rhymar. I am here with Manny today. We are broadcasting live from Foster City and today we’re going to talk about the future of knowledge work. Uh Manny and I have both been in the space for a while now and I’ve kind of seen this evolving. Obviously, Replet is leading the charge in lots of ways, but today we’re going to talk about generally knowledge work, how you can get the most out of uh your AI prompts. Uh we’re going to have some replet demos here that we’re going to dive into and kind of explore uh what you can do

01:41 inside the agent chat. How you should be thinking about your knowledge work moving forward. We’re going to have I think an interesting conversation generally about how the world is transforming in front of us in real time. So that’s going to be exciting to me. Uh Manny, any thoughts as we get started here and then we’ll shout out a bunch of folks in the crowd. What’s up everyone? >> I am excited about this Rimar. I I feel like, you know, even people in the space right now are underappreciating how much their workflow is going to change. We’re seeing it with some of our

02:11 power users, what they’re doing with tools like Replit internally, what you’re doing. And so, even if you’ve been plugged into the AI space, I think you’re going to walk away today with a new appreciation of a what you can do today and b where the puck is going, how things are changing. >> So, I’m pumped about this. I think this is going to be a really good episode for everybody. Likewise. So, let’s give a shout out. What’s up, Dario? I see you in the chat. Hello. Hello. We got Brian in the house as well as uh Alva. Hello. Um Rashir, hey, what’s up everybody? Uh

02:45 Pavll in the house. Ker, what’s going on? Hello, Ruth. >> Alden here. >> Yeah, shout out and let us know where you’re watching from. We always like to see that in the chat. Let us know if you have any questions as you go. And specifically, we’d like to know what you’re thinking about the world of uh kind of knowledge work generally. Uh you know, this is um evolving in real time. We talk about the speed uh how fast this is like moving right now. And so I think um what we’re going to do, how the show is going to break down is we’re going to

03:15 have a conversation about what is knowledge work. Uh we’re going to go through a little bit of how the space is evolving. We’re going to then do some demos and then as we typically do we’re just going to have a nice conversation with you all with the crowd uh to discuss some of this and we’ll just get into the weeds um as we build with the agent and kind of discuss the changing landscape. So uh that’s that’s essentially the plan for today. Um okay quickly any community updates. So again today we’re doing this just ask replet episode. We’re going to talk about the future of knowledge work

03:47 tomorrow in office hours. if you’re not uh around or if you haven’t been to an office hours, we’re talking about pricing your replet projects. So, this is something that we have discussed uh a couple times kind of around the edges, but we’re going to get into the weeds. Um so, if you’re freelancing and thinking about, hey, how do I price my uh product? If you’re an agency and think about how do I land bigger clients? If you are thinking about like the different pricing models, you know, you could uh fixed cost um you know, uh time and materials, uh project base, like we’re just going to kind of explore all of that tomorrow in office hours. Um

04:20 we’ll also be on with the reream folks. Many you want to um share a little bit about this, but we’ll be doing a nice little uh interview tomorrow with the folks at Reream. >> Yeah. Yeah, for context for everybody. So, we use Reream as our streaming platform to get the show out to all of you. And the cool thing about Reream is they have some really nice APIs in an MCP now that allows you to custom build tooling. So, this lower third that you’re seeing here at the bottom, the nice peel at the top. This was all built by Rhyar. It’s being brought in. He’s got some other tools in the back that are working and helping us run this

04:50 show. So on uh Reream’s live stream tomorrow, we’re going to be talking about how we’re leveraging their webinar functionality and how we’re building out our own custom tooling that sits on top of Reream. So going to be a fun show. I’m looking forward to sitting down with the pros and talking about streaming and webinars or gated live streams and the difference between a webinar and a live stream and the different vibes and how do you how do you build your own custom tooling on top of a platform like Reream. So, it’s going to be interesting to to have a chat with them about that.

05:21 >> Yeah. And then as always, we got the Friday showcase coming up uh you know this week and we’ll have some fun interviews and community showcase things to cover. So, action-packed week um talking about knowledge work obviously pricing your replet projects, the reream interview tomorrow and then Friday we have this community showcase. So, lots of fun stuff coming up. >> And just to call out again the office hours, I’m really excited about that one too, Rhymar. People really seem to love these conversations where we talk about like how to make money with what you’re building in Replet. A couple weeks back we started this conversation and we ended up just talking about uh ads uh

05:53 leads top of funnel type of stuff how to build a community and we said hey we you know this is like a rich topic let’s come back and explore it again. um really good conversation and the cool thing about those office hours is it’s not just me and Rhymar yapping like most of it is like our power users getting on stage and like just sharing everything in terms of what they’re doing and I love it because like you’re you’re learning from them you’re exploring builders and what they’re doing um and so highly encourage you to to check out our office hours tomorrow and I’ll drop

06:23 a link to our Discord in the chat as well for you to go and sign up on our Discord. Yeah. And um again, shout out in the chat. Uh I see a lot of the regulars, Ruth and Nadia and Non jumping in here. I see Gatis in the house. Uh I see uh Daniel Kim here as well. So yeah, thanks all for hanging with us. Um let’s jump into the show. Let’s jump into the presentation today. Um okay, let me go to this screen and then to this screen. All right, we’ll bring Manny right back

06:54 here in a uh few minutes. I’m going to run through just a couple thoughts I’m having about generally um the future of knowledge work and setting this up. So, my name’s Ryar. Uh I have been doing knowledge work for over a decade now. Um building websites, doing all sorts of different things and kind of watching this space evolve in this creator sphere. And what I’m seeing right now that excites me is that these tools allow us to amplify our efforts, right? And so, uh, you hear this term slop cannon. You hear a lot of AI slop stuff.

07:26 Um, and this is surely true, right? Like you can create sloppy things and not great experiences. Uh, if you’re kind of using these things to amplify bad outcomes, but what we’re going to see here, and what I think we’re headed towards is this world where like if you have a little bit of taste, if you have a little bit of insight, if you have a little bit of like uh quality control, you can really amplify the things you do with, you know, uh, with these tools. So what is knowledge work? Knowledge work is work that uses knowledge and judgment to turn information into decisions, ideas or solutions, right? So analyzing

07:58 data, designing a product, writing a strategy. This is not new, right? Like knowledge work is not anything new. We’ve seen and had knowledge work for a long time. Uh and if we go through this idea of like what work is generally um we have kind of come through a world where like almost everybody was doing agriculture and this is like you know we’re going back in time, right? So uh everyone was kind of focused on physical labor and survival and like this was the thing that we all had to do. And then slowly we had this industrial revolution and machines came along and amplified our physical labor and we started seeing

08:28 people move then into like office/service/ less physical labor type stuff more service-based things and we started to then see in the computer space automation supporting that routine work. Uh and then over the last handful of years as the tooling has gotten really powerful, you’ve seen this whole idea of knowledge work kind of expand exponentially. And the idea is that you have knowledge as a human uh or as a business, you have some kind of deep unique insights and you are taking those and turning that into actionable things. You’re solving problems. You’re applying that knowledge in an interesting way.

08:59 And obviously now we’re entering this AI powered era where you can amplify the impact of what you’re doing exponentially. And if you’re a tastemaker, if you have high quality, if you have like high agency, then this is going to amplify those efforts and allow you to kind of do more. But also, if you’re doing skezy scammy stuff, you can also amplify those types of things. So, I think we’re entering this era where you as an individual are going to be the difference maker when it comes to what knowledge work is. So, the idea of knowledge work is is evolving. AI is

09:30 accelerating execution, but you still need to bring that taste and enforce the quality, which I think is the the the big takeaway from a high level here, right? And so, as we’re thinking about this, as we move into the demos in the next phase of of this video here, I want you to start thinking about like how do you approach your knowledge work? How do you prepare the things that you want to do? And what are you thinking about repeating? We’re uh I’m sorry, um delegating and automating, right? Because we’re kind of entering this era where you can build anything. We’ve seen this uh and we’ve seen a lot of people

10:01 build these like crazy situation hubs where they’re monitoring the situation and they have all these inputs. But what I want to take away from what we’re going to talk about today is that like thinking about small pieces of this, right? Like thinking about how to delegate the repeatable task and focus your attention and like use these tools to really amplify your taste, the thing that makes you unique, the experience, the emotional connection with the user. Uh you know, like is is is how you measure those outputs, right? That the outcome, the thing that you’re trying to

10:32 get at is like, you know, this human bucket that is in the loop of what you’re trying to build and then the AI comes around the edges to amplify that again to build on what it is that you’re thinking or how you’re approaching this thing. And so, as we again go into thinking with AI and as we think about knowledge work, you want to start thinking about first principles. We talked about this in office hours the other day and especially as you’re thinking about your prompts here. I’ve got like two more slides and we’re going to go into like thinking about prompting and thinking about how to approach the

11:02 chat, right? And so a lot of times, you know, you might hear or be in a situation and think like uh this is a perfect example, we need a weekly report or we need this thing, right? And so the high level, the urgent need, the fire in the moment is we need a new weekly report. And I would encourage you that as you’re thinking about knowledge work, as you’re thinking about the things you do, as you’re thinking about the value you bring to the table, is instead of just saying like yes, let me go create a weekly report, think about what is the underlying ask that is there, right? Like is this your boss asking for some kind of uh alerts because maybe they

11:37 need to know when something changes, right? So imagine the weekly report is the ask because up until now the only way to get the insights is to wait until the week ends and then create the report and be able to look back on what happened. Whereas maybe you’re in a moment where you could create a system that instead of waiting for the weekly report creates a real-time alert that notifies you in that moment and says, “Hey, we need to adapt this right now.” Right? Right. And so that’s kind of the change, the moment that we’re at in like knowledge work where it’s like you used

12:07 to kind of have to wait for this thing to take place and for things to happen and for you to like iterate and then make the change. Now you have the ability to kind of build a real-time feedback loop, right? To to kind of take the impact or to take the data to take the thing that you’re trying to do and really spin that up in real time. And so maybe the ask is, hey, create this weekly report, but your idea should come in and say, hey, what are we trying to accomplish with this report? and what are the real requirements that we might want to build here that could solve this problem in maybe a more efficient way, right? And so this comes down to

12:39 thinking at a higher level of abstraction. This is another thing that as we talk about knowledge work, we talk about becoming a systems thinker and thinking about these processes and if you’re going to automate chunks of work, if you’re going to kind of build uh proper systems, you have to start thinking about this thing holistically, right? And so again, as we’re going to talk about prompts here in one second, I want you guys to think about like what is a basic prompt? You know, we’re going to look at like what is a basic prompt, what is an intermediate prompt, and then what is an advanced prompt. But I also

13:10 want you to start thinking about like how do you step back again from that first principles concept. Where do you step back and you look at the whole system? How do you look at this whole piece as one thing and say maybe they’re asking me for a weekly report over here, but what can I do over here in this system to like anticipate the needs of the report or to eliminate the need for the report like we said with like a real-time alert that then allows you to take another piece of action, right? And so the whole concept here as we talk about knowledge work, as we talk about automations, as we talk about bringing

13:41 AI into this world is looking for little pieces where you can optimize the system. It’s not looking for these big chunks, right? Like I have this quote at the bottom of this slide. I don’t know if you guys can see it. Let me hide the lower third here for a second. Uh and it basically says to to improve is a change, to perfect is to change often. I don’t know if it’s like a Chinese proverb, an old quote, whatever, but like this is kind of the idea and this is something that uh is really strong inside of digital systems as well because you have this iterative loop, right? And so as you think about knowledge work and as you think about

14:12 you know how to apply your real- time u learnings into like a space in a practical way it’s not just this idea that now I have the knowledge and now I have the tools now you have to step back one layer beyond that and think like okay knowledge this is me this is what makes me unique these are my inputs here’s the system here are the decisions we’re going to make and then again here are this here’s this loop we’re going to build that we’re going to check these outcomes and at every moment we’re going to adjust these knobs and levers and this is what we’re going to like this is the idea of thinking as a system rather

14:44 than like thinking of like okay we need the weekly report that’s just a a set outcome that’s like a deliverable of the system but what is the system that you need where are you collecting data for the weekly report what are the alerts that need to come into that weekly report what are actions are being taken from that weekly report right and so instead of just thinking hey weekly report you have to think about here is this system that’s delivering this output which either a weekly report or if we look back on the last slide maybe it’s again some other type of of of

15:15 metric or alert that tells us you know uh what we want from that weekly system from that reporting system right okay so why is replet leading the charge in this space uh why am I excited to be here as a creator because if you don’t know I kind of came in from this space just building on the side and now I’m building here uh what’s up Zack I the uh pineapple on pizza the debate always lives on. Okay. So why is replet leading the charge birth in industry vibe coding? So um

15:46 over the last year or two you know we have pioneered this concept of vibe coding. Uh obviously you saw the whole industry evolved the infrastructure that uh you know agentic software development. Replet was one of the leaders in the space. But one of the things now as you think about knowledge work and one of the things that like really is a competitive advantage in the world of Replet is this idea that like we’re not tied to any specific model, right? So if you’re using a codeex, if you’re using a claude code, if you’re using a Grockbot, you’re kind of tied into their like core model where in

16:17 Replet you have the ability to kind of like dance between the models. And this becomes really interesting because now we have the ability to kind of orchestrate and hand things off to different models that might be better at design or software uh uh auditing or maybe security or uh you know whatever the other the pieces as these like technologies evolve. Um you have this again the ability to like inner inner uh jump in between them and then you have uh the ability to take an um an idea from this concept to

16:49 production. I want to share uh a little story here in a second about a friend of mine who built something uh with with uh another AI tool recently and now they’re kind of stuck in a spot that like leaves them really tight. But we’re going to get back to that example here in a second. Um so Replet has the best agent harness. Again, we talked about uh the ability to uh transition between uh um the models and then we’re just the product’s leading, right? Like we’re just catching up on the marketing side, the communication side, etc. Uh the product has kind of been leading the charge. And so what I want

17:20 to do at this point is jump into the idea of shaping your prompt, go into replet itself, and show some real examples here. So, um I think that’s just the the entry level idea of what knowledge work is. Um Manny, I want to share an example uh of a friend here um that that um I I just just chatted with this weekend. But before we get into that, any anything I missed any uh thing from just like the baseline of what knowledge work is and how people should be thinking about this before we get

17:51 into real demos? um any anything coming in from the chat or thoughts that maybe I missed that that would be good to hit here before we go any deeper? >> Yeah. Um right out of the gate, I really like your framing around like not just doing a task, but also exploring why are we doing this task? What’s the value? Because what I’m hearing folks with that are using AI is that it’s allowing them to automate their tasks, but a job is just a set of tasks. And so if you can quickly complete those tasks, you can move up the value chain, so to speak. So maybe you were the person that always

18:21 did the copy on a tweet or on a blog post. Now you can leverage AI to do that better and and it frees you up to think more strategically in terms of how you’re going to position a product or service or how you’re going to go to market in a different way. Um it also allows you to be more full stack than you were before. We’re seeing this all the time with designers and PMs that are now leveraging AI. A year ago, a PM’s role was to put together a product requirements document to list out all the things that a product might have. Today, I think the expectation is that you’re vibe coding the first version of

18:53 that application and you’re putting it in front of users and getting real-time feedback. If you’re a designer, same thing. It used to be that you just did the front end, right? All the pretty pictures and everything and then you threw it over the line to the engineering team. No, now you can vibe code that and show that to folks and get feedback on that. And so I think it just allows you to be more broader in general. Um the other thing that comes to mind, Remar, is just kind of this idea of the 10x engineer. And I think engineers in particular have are seeing this first because the real kind of feel the AGI moment hit in December with

19:24 models got so good at writing code that engineers overnight woke up to it and say like, man, like I cannot function anymore without these tools. And if I’m doing it right, I get to 100x my output. And I think that’s straightforward for an engineer because all of it is code. But more and more every role if you’re a knowledge worker is going to go through that process as well. You’re going to be able to 100x your output. And so your job I think is to figure that out. What is that value stack you need to move up to up up through? What is that full stack of capabilities that you have to bring to the table now? And how can you

19:56 100x your capability? Because the chances are that your tasks are probably going to change and streamline but your job isn’t. and it’s going to have the kind of the same kind of overall kind of feel to it but just doing more. Um the case study here for me is around X-ray technicians. Like a year ago, two years ago when AI was just popping off, everybody’s like there’s not going to be any more X-ray technicians anymore. It’s going to be done anymore. There’s more X-ray technicians than ever. They’re not spending the same amount of time looking at all of the scans and like reading all the detail. They’re leveraging AI to do

20:26 that. It’s just the nature of their job has changed. And so the tasks have changed, but the role has evolved with AI. >> Yeah. And I again I I and I I agree 100%. I I’m not one of these like knowledge work doomers AI like is going to take all the jobs. I actually think it’s the converse of that. I think we’re going to create more jobs than we know what to do with for the people who have agency. And again we’re seeing here early adopters can competitive advantage by kind of thinking through some of this stuff. Um, I want to talk uh two two two

20:57 examples um about, you know, uh first principles thinking that relate back to the show here, Manny, and we can go on a little u trip down memory lane because when I uh started, one of the first things you asked me to do was to like copy and paste manually all of the chat out of these live stream comments, right? So like at at one point you know the the question was or the ask was so I I want to you know going back to this we need a weekly report type of thing and what is the real requirement and how you guys can think about some of this. So so

21:27 the the immediate pain point was hey we have these YouTube live streams that are happening. There’s more chats coming in in real time than we can kind of address. We would like to make sure that we pull all of those chats in so that we can take that feedback and action on it even if we can’t address those in real time on the chat. Right? And that was kind of like the thing, right? And the process for that in the moment was manually copy and paste all of these chats and then at the end put those in some kind of spreadsheet and then run them through some kind of AI, right? So

21:59 they asked Rhyar to do this thing. Rhymar is deathly allergic to uh repetitive like work. And so I did it one time and then what I again my thinking was the task is irrelevant, right? like how these chats get into the thing like doesn’t really matter. This is just the way it’s happening now. The real output is like we want these things into a system. And so created a little browser extension. It scraped the comments. Uh that system has evolved now uh into something that we use uh regularly for the shows that now pulls

22:31 in from Discord and from chat. And now it kind of automatically processes these episodes at the end. It pulls the transcript from reream at the end of the episode. uh tune in to uh Wednesday for more details on how this actually works. But again, going back to first principles, the idea was okay, here’s how things are happening now. The ask is just do this thing. And I could have probably just kept doing that thing and you know, somebody in some world would just have like done that and that would have been a task that a person would have done all day long. Maybe somebody would have figured it out. But like no,

23:02 no, we said, “Hey, what do we want here? We want this process. We want this thing done at the end of the time. How can we automate that? Okay, go get the chat in real time. Build this browser extension. and we use Replet to build a little browser extension, feed them into the back end. And now this whole process happens on an automated way. It even generates the draft of the email that we send out to recap the show or the post that we create in Slack to update the team about what happened that day. Uh, and all these other ancillary things. So again, it’s taking what was a couple hours worth of knowledge work from our end and just within 20 minutes of the

23:34 show ending, it’s done. Um and so that’s you know one example I think where you can kind of like take again what was the ask totally different from what the end result was but if you can step back and look at these things like that you might find that you become the person at work that like finds all these little efficiencies uh you know to take again things to the next level so um >> I think it’s a great example a couple things one it’s so much more joyful to build an automation it is to do that like mind-numbing like copy pasting thing uh two is it frees frees up our

24:06 time to focus on more higher value tasks. So if I’m not spending 20 minutes copy pasting and also it’s error pro prone so like your system is like 100% whereas like somebody like manually copy pasting that more errors get introduced to the system and then three you don’t know what additional value you can start building upon it. So like now as you mentioned like originally the task was like we needed to to save these comments for posterity but now you’re adding analysis on top of that. you’re seeing trends and so you have this really cool tool that our PMs can go to and say like

24:38 hey tell me the times when people when somebody mentioned a security agent issue or like this about this feature what people like and there’s some comments here in the chat like somebody there’s some feature request stuff like a PM can come and now find those so like you’re creating more value on top of this >> yeah and what we’re looking at the screen now is basically this chat tool that I’m talking about so if we go to all messages and as you guys are sharing and saying here are the chats coming in. We can uh filter them by platform. Let’s say we want to just see what’s coming in from X. Let’s say we just want to see

25:09 what’s coming in from LinkedIn. So, this is um you know the messages here. And what it does is it blocks them all in by stream. So, we don’t have to go into the weeds of this, but I just want to show you again the things that started out as a little thing that we’re going to jump into. Uh maybe we should, you know, like jump in and throw a couple prompts in here. um and get the we’ll get the agent going in the background so that we can show a few demos and then we can you know go uh wherever we want to go. Um and then I have again an example that I want to share uh with with folks about a friend who um it’s a bull signal but

25:41 also uh a reason why you should replet uh use replet. So um let’s go through this thinking through your prompt. So okay so we’ve kind of now laid out this idea of what is knowledge work how this thing is evolving and I want to kind of like show you a few things you can like let’s just talk about how we approach prompting or how we approach kind of using and getting the most out of replet uh Manny so um okay so a single source prompt right like if you’re just kind of getting started you might approach the world of let’s say you use a granola or

26:12 a meeting notes tool and you’re like well I don’t need replicas I can just talk to in granola or I can just go to chat EPT or I can just use you know like this other thing right sure 100%. You get then a single source prompt with surface level insights. You might have a prompt like take a look at my last week of granola meetings and give me a high level overview of what I should be working on today. That is like an example of a prompt you could put into replet for sure. And you could connect the MCP to replet. Actually, if we look over here, I have a very similar prompt that I asked and I said, take a look at my last two weeks of granola meetings and give me a highle overview of what I

26:44 should be working on. And we can see here that it kind of turned this, you know, out and gave me information about what I should be working on. and I said, “Give me the full rundown.” And then it gave me more. Right? So like this is something you could ask, but then you could also just go to Granola and ask this. And so you’re probably looking like why is this unique? Why do I need to, you know, think about replet here? Okay, so that might take you to some kind of intermediary prompt, right? Or intermediate prompt, right? Something that’s a little bit deeper. So again, as you’re thinking through the prompt, maybe instead of just saying, “Take a

27:14 look at my last week of granola meetings and give me a high level overview of what I should be working on today,” we want to say something like, “Take a look at my calendar over the last week as well as my Slack conversations and granola meeting notes from the same time frame and give me a high level overview of what I should be working on today.” So I could take this and drop this into a prompt here. Let’s see. I think I have this set up. Nope. Got one more. Okay. So, I’m just going to send this, copy and paste that, and just drop it into the mix here. And so, that might be like an intermediate prompt, right? Um, but

27:47 and this is something that I might have to do on a regular basis. The advanced prompt or something that you might think like, hey, let’s just like optimize my whole world and make sure that I’m ready every day for, you know, what I want to do is a prompt like this. Every night at midnight, I want to take a look at my calendar for the past couple days and the next couple days of meetings as well as my Slack conversations granola notes and notion docs from the same time frame and give me a daily briefing highlighting key projects and what I should be working on that day. So, and then I’m going to this I say advance

28:18 because I’m going to ask it to do a routine. So, we’re going to do two things. I just asked it here to kind of like look up this calendar, Slack, and granola notes. And then in this other window, I’m going to basically give it instructions to do something on a reoccurring basis here. Right? So that’s the routine we just set up. Every night at midnight, I want you to take a look at and then I’m listing these things and then telling it to give me a daily briefing. So basically, I want to wake up to what should I be doing today, [laughter] you know, like like what should I be like, you know, obviously I

28:48 keep track of this stuff, you know, but like let’s see what this thing comes up with and we’ll see how the agent kind of starts or how Replet starts thinking about working through these weekly priorities and what we’re actually going to like um get accomplish here. And one thing I want to point out is that I do have uh these integrations kind of already set up. Um so where where you’re seeing me uh look here is on the left side of the UI you have these integrations. When you ask the replet

29:18 agent uh it will in line kind of ask you to um authorize these integrations. Uh and so um I didn’t disconnect them because I don’t want to reconnect them all. But if you look on the left side of your UI here you can see your integrations and you’ll see all available integrations. So there’s a ton of different integrations. You can come and add uh any of these. You can add a custom MCP. Um and then if you click on your integrations, you will see the different things that you’ve connected. So we can see here that I’ve connected

29:49 uh Gmail and my calendar and notion and granola. Um and I’ve given it permissions. You can kind of set the permissions that you give as you do this. So you can set it to always ask for permission to decide for you or to always allow. And again, each of these integrations can have their own uh permission for how they approach or do uh different things. It will also then show you which applications are connected to those and what the status of that integration is. So, just a heads up as you do things like this. Um

30:20 it will it will try to connect different things. So, okay. So, power is a better fit for this. Okay, great. So, I’m going to go ahead and continue on power. Uh we started on free. Um, and we’ll see that every once in a while the agent will ask for a little bit more juice. Uh, especially if you’re going to try to do a lot of things like this. I’m asking it to look at Slack, Granola, and Notion uh for the same time frame or I’m sorry, in my calendar. Uh, so that’s like a lot of stuff to to try to do in a small context window. So, the power will give it a little bit bigger context window. >> I think this is super helpful because sometimes I wake up in the morning and

30:51 like like I’m trying to like get stay off social, stay off Slack. There’s so much going on. And so it’d be nice to have like a nice summary that I can look at first thing to come just help orient me and keep me off of those those bad habits. >> Yeah. And I think um again, no matter how good you are at organizing your world, there’s still things you’re going to miss. And so even if this doesn’t become your like source of truth, the ability to like have something like this to just at a glance remind you about what’s going on. >> So let’s kind of look at what’s happening here, right? So um >> all right. So I’ll set up a daily schedule briefing is what the agent

31:22 says. Um, and this is one of the things I want to point out. Like a lot of these other platforms, you got to like set up your chief of staff and you got to set up your marketing bot and your whatever thing and then like orchestrate like with Reply you just come and ask the agent, right? Like you don’t have to worry about the personality of these different things or setting up a bunch of rules or a bunch of like instruction files. like there are things that are in, you know, like it’s just going to kind of do the work, you know, like uh I see lots of excitement around all these other AI tools, Musebot, Grockbot, like

31:54 all these different things that are kind of like trending virally right now. Uh they’re great tools. I, you know, go explore them. One of the things I love about Replet is that I don’t have to worry about like a lot of those small setups or different things that you have to do inside of there. And so, um let’s look at what this thing did here. So again, as a reminder, I’ve asked it every night at midnight. I want you to take a look at my calendar for the past couple days and next couple days of meetings, as well as my Slack conversation, granola notes, and notion docs from the same time frame, and gave me a daily briefing highlighting key projects on what I should be working on that day. What it has created here is

32:25 what’s called a routine, right? So, right in the chat, we’re going to get a routine. And this is basically taking what I told it, it created a prompt here. And what it’s going to do is on a reoccurring basis at 12:00 a.m. every day in this time zone, uh, it’s going to give a daily project briefing. I’m going to go ahead and create the routine. And we’re going to watch the UI do a couple things here, right? So immediately it’s going to give me try a test run so that I can like run this thing if I want. Um, we can also see on the side here now that I have a routine kind of set up, right? And so this gives me an overview

32:57 of the routine. It gives me whether it’s active, what the budget is on the sideline here. I can decide how much or how little I want to like cap on the spend here because this is going to spend some bucks. Even if you uh are in free mode when you create the routine, the routine’s going to run in um either power or max mode. Uh so just a heads up, you do need to set some kind of budget here. And what this will do is make sure that if you set up some crazy routine, it just doesn’t like you know go spend all your money at once. Um and it shows you the prompt, what it’s

33:27 asking, etc. So let’s see. Your daily project briefing is up. It runs every day. I’m going to hit test run. Let’s see what we get from this thing. Uh, no assets have been created. No integrations. Let’s see. No integrations used yet. Uh, this will show up after the first run, I do believe, once it hits all those things. Um, so this and and I think what we’ll do after this is we’ll just have this also send me this report in Slack. Um, after this is done. Um, and then what we’re also going to do as the next demo is explore some data. And then if you

33:59 have thoughts or ideas for what you’d like to maybe see or experiment with, and you want to drop those prompts in the comment, we could uh dive into some audience suggested prompts as well. Um, anything coming in from the crowd that we should um take a look at while this uh first run is is happening here in the background? >> Yes. Yes. There’s a good actually a good question that came through. Let me go to pinned here. Um, you’re integrating everything into your existing apps, which is cool, but there’s a concern around privacy. Can you work with um like can I use replet work without

34:31 having it access databases or other sources? >> Um, yeah, you so the only thing is it’s going to just be limited by the the context you give it, right? Yeah. So, if you um again going back to like these prompts, right, these beginner prompts, right? So if you have like a single source prompt then you’ll have the context of just your granola meetings which may be everything you need right um the the the the good part about giving it access to these other tools is that it has like a whole look so here we can see in this um

35:01 in this uh um routine here now we can see Slack rep uh granola Google calendar and notion are all connected here um and so you know like the more you give it, the better it’s going to be at kind of giving you a better view. So like yes, you can accomplish some of this stuff with limited access. Um, one of the cool things with uh Replet and one of the I think emerging uh benefits to working on Replet is the enterprise rails and security and kind

35:33 of the the sandbox with which you operate in. We hear a lot of like these AI companies talking about sandboxes and AI’s escaping things and data leaks. And um again, one of the things that sets Replet apart is that you have this fully integrated environment that has robust security protections. The connectors are like kind of inside the same sandbox. Your database is directly connected, you know, to this. So you’re not bringing in like an external service and you’re not relying on like uh external connections. The database is kind of married to the

36:04 agent. Um, you also get the separation between production and development, which is a major thing as you’re building software. It’s not as important when you’re doing something like this because this is not like deployed. And maybe we can talk about the difference between, you know, like deploying an application versus like working in the agent chat like this. Um, but yeah, you you know, you you can accomplish some of this. You’re just going to have limited results with less context and less access to the data and that type of thing. So, um, >> and I would say the connector, I would

36:34 go with the connectors. So, like if it’s a connector, we’ve built that out really well, uh, to to do best-in-class privacy measures. Um, and even when you go through the connector, there’s some steps that are in there so that it manages your privacy in the context. Like, if you use the notion connector, it’ll tell you specifically what pages you want that replet project to have access to. So I would encourage you to tiptoe into some of those connectors start leveraging them because uh to Brimar’s point uh you’re leaving a lot

37:04 of context on the table like everything that’s happening on Slack and notion your inbox your calendar you want to bring that in so that replet as you can see that and give you like a really nice summary that’s useful. >> Yeah. Yeah. And um so here it is the calendar shows it just as replet session this morning. So, it’s kind of looking through its thing. And again, we’re watching this happen in real time, but the idea is that this would happen overnight while I’m sleeping. And every morning, I would just kind of wake up to this already prepared for me. Um, and so we’re just kind of uh we say do it live, you know, like we did.

37:34 >> Yeah. Um, and let’s go let’s go back over here. Okay. So, so, okay. So, two different things we asked, right? Uh, right here, we’re building a a routine. this. If I look over again in my sidebar here and I click over to routines, I’m going to see that I have two routines here. One is my test prompt and one is the the one that we were just building. So, I’ll go back to that. And we always have this routines on the left side here of our sidebar. Um, and then if we go to

38:06 this other one, what we just did was ask it in the chat. So, this is not a routine. This is just, hey, we asked it again. This is just a basic prompt. We’re here in free mode. Um the broad Slack search returned. Okay. Today’s priority is live content and tomorrow’s prepped. Your calendar has a just asked replet deep dive community marketing sync at 4 with a deep workday blocked around. Then uh run the deep dive protect a work block. Use the 4pm sync to unblock. So I could dive in to different things that are happening here. One sale item to cross off Monday’s notes. Okay, great. Um yeah, so

38:37 again, this is just a quick snapshot. So if I was in the agent and I just wanted it to create the snapshot for me, that’s there. And then this is the daily summary which should be fairly similar. Uh but this is going to run daily for me, right? So uh let’s see. Resolve tomorrow’s reream timing and conflict. Interesting. Uh draft the pricing outline. Use the 4pm community marketing sync project polls. Today’s show was reported schedule on reream, but the team did not decide to Okay. >> So [laughter] that’s pulling a lot of DJs. >> It’s pulling a lot of stuff here. Yeah. Yeah. So again, this is something that

39:08 um I I would probably tune in, right? Like at this point now I have the basics. I might start telling it like, “Oh, hey, I don’t need things from this database in notion, so leave that out.” Or, “I want you to focus on these channels in Slack, so take a look here as the primary source for the data.” Um, and so this is where again, like you just come, you start, you get, you know, we talked about that iterative cycle. You get the first loop at this thing. And, and you see here, like none of this is software, right? Like we haven’t created an app yet. None, like we could

39:39 turn this into a dashboard. So, we could theoretically turn this into a dashboard after we tune this in. So, I might let this thing run for, you know, a week or two and start tuning it daily to start figuring out how do I like lock this in to pull in from exactly what I want and how do I like maybe give it additional context as we go. And then I might turn this into a full dashboard or I might turn this into some kind of tool that gives us a snapshot of what’s happening in the space. Um, >> yeah. >> Um, can I do an important call out here?

40:10 This is Dario. This is an important point here. Yeah, the nice thing about routines is that you don’t have to be logged in for them to run. So, Rhymar is setting it so that it runs at midnight, does all the work, and then it processes the report and it’s ready for Rhymar when he wakes up. This is the key thing. This is a nice thing about Replet in general is that it’s cloud-based. And so, if you’re running some you’re if you’re working on something in Replet, an agent is working on something, you close your laptop, you walk away, that that process is still working out in the cloud. That’s something that differentiates us. Um, but then routines

40:41 takes that to another level because it’s scheduled like you don’t even have to log in and to and to run the the job. It knows to do it at a regular cadence. >> Right. Right. Uh, and the uh uh to tag along to that is that you don’t even need to worry about deploying this thing. Right. So like when you’re building applications or when you’re building software for other people to access it, you often need to think about like deploying it or having a URL or having a place for people to go and interact with it. This is a totally different experience. When you’re in the chat, when you’re kind of doing this type of thing and when you’re creating these

41:12 routines, you don’t need to um worry about deploying anything or having a server running all the time like Manny was saying. This is just going to kind of run at your scheduled increment. This could run hourly. You could set it to run daily. You could set it to run, you know, once a week, however you want. Um, and so this is a real good way, like going back to like create a weekly report. This would be a real good way to start creating that first version of your report is just to come give it the sources you want and say, “Hey, create this thing for me.” And then from there,

41:43 you can start kind of honing in on what you want that thing to become. and slowly kind of tweaking the prompt or tweaking the the routine till you get to the point where you’re like, “Okay, now I can kind of fully automate this thing.” Um, and maybe you then want to build a piece of software or something that you uh can deploy that just runs all the time rather than like on a routine or in the background. You can just have the system running for you. >> And one um another question from that’s right. So what Rama showed you earlier

42:14 when you set the budget that’s the limit for what that run is going to cost you. So don’t mistake that with the cost cuz like the cost is usually like way below that. I think it’s like cents for like running a an easy routine, but that that budget is the maximum amount. So that run is not going to go over that budget for you. So um that’s right, Mam. >> Yeah. And look here 67 cents. Uh we worked for 5 minutes. So basically what you’re saying is like here’s the max we want to spend on the routine, right? Okay. And so if I open this uh on the

42:44 sidebar, you’ll you’ll see, you know, again, like you have the agent chat and then you have in the top right, you always have this ability to access the sidebar here. It’ll show you any assets that are created and again what integrations are being used in this. Um and you could always, you know, view this prompt. You could ask the agent to modify the prompt. So inside of here, I might start telling it because this is where going back to like the knowledge work and where the human taste comes in. This is now like a very rough version of what

43:14 this report thing should be. From here, I would start thinking about how to lock in what I want to give it as context. Right? So now it’s kind of got the whole window of my possible work world and I might want to say, “Hey, only pull from these Slack threads or only pull from conversations in our community HQ channel.” Uh so that we stay focused on like community stuff. Um, and so from here is where this is again where the human in the loop comes in where you start iterating and start tinkering with

43:46 what you actually want to get out of this thing. So, >> you might break this out by projects too. You might say like for this recap like I want you to update pull from this notion database. Check these Slack channels. Check for these keywords in my email and like give me like any updates on this project. Particularly helpful when you’re like you have a project that maybe you’re waiting on other people for things on and there’s like maybe like a one two day window. So like kind of forget about it. So I you know it’d be nice to set up a routine like yeah like let me know whe there any updates here on any of these channels or sources and

44:16 like let me know. >> Yeah. And I think that might be again going back to a more specific prompt right if we talk about the beginner versus intermediate versus advanced. Imagine a prompt that said something like, uh, I want you to check Slack every hour for me, and anytime it mentions this project, >> create an alert for me or bring or put that on a list on a on a on a bulletoint list and and add it to a a notion database,

44:46 >> right? That way, instead of it just pulling every possible thing, you have six routines running, one for each project. And as these things happen or as a calendar mentioned or as a person let’s say hey anytime this person uh makes a post in this channel in Slack I want you to bring that to me >> right and so that’s where I go back to like this is you know just getting started you have to kind of think about like what’s the first thing I think a lot of times when people you know uh approach an AI system they’re like what do I like what do I prompt this thing

45:17 what do I actually ask it to do um and so sometimes experimenting with the data connections experimenting with what you ask for uh you know and just kind of like getting a starting point is where you want to where you want to be and then once you get that starting point then you can kind of shape what you want that to to become right like then you can start thinking through like what is my system prompt or what is the next thing so um maybe we can if we have questions about this routine we can answer those and then I’m thinking um we have another demo that we could jump into here uh to get into some like data

45:50 analytics and how you might approach uh doing some analysis on a large data set as the next step. >> The only other question to call back is Nadia earlier in the show saying that she implemented a morning business update routine and she’s curious about how to take it to the next level. So hopefully Nadia, you’re getting some tidbits here on some of the demos that Rhymar is showing you. >> Yeah. Um and maybe I can do maybe we can use this chat scrape again for like what um what the next level of something like this may be, right? Because if we talk about like if we go into uh this episode

46:23 for today, we’re going to see that it’s got like an incomplete data set, right? Obviously, the YouTube analytics are not in yet. The performance over time is not in yet because the stream is still currently in progress. And so, it’s saying right here, hey, this is live right now. Uh, you know, we don’t have the data. Same thing with the insights, transcript, chat, or a Slack recap, right? We don’t have anything here. But what I have on the right is this system prompt and an editorial default, let’s say. Um, and so if I go back to another episode, and again, this tool started

46:53 with us manually scraping comments and dumping it into a spreadsheet and then running that into some AI service after the streams. And so it basically started like this routine that we’re creating here where we just wanted like one thing to kind of happen and then as that thing happened, right? So like we have a different example here where like the chats are coming in, but now this thing is fully automated where a as the chat ends, right? So now it draws us a little curve performance over time. It pulls the YouTube analytics in for us so we don’t have to go hunt that down in

47:24 YouTube. U it pulls the reream data for us. But what it also does is it processes the chat and transcript, right? And so these things that were manual before. So Nadia, you’re um talking about how to like take it to the next level with the routine with your with the way you chat with the agent. That’s you just want to get base level, you know. Okay, this is step one. And then once we figured that stuff out, we turned this into like real prompts that like just do the things for us on a repeat repeat basis and put it into this like dashboard for us. So now anybody

47:55 can come and see, hey, what are top questions people are asking? What’s product feedback people are talking about in the chat? What are the support requests that people are making? What are notable quotes? What are content suggestions? Things that we may want to like, you know, consider for future uh episodes. And then who are the uh top contributors kind of on a per episode basis here, right? And so same thing with the transcript. Here’s the insights. We can copy this if we want to use this as an email. So as you build your routines, as you build pieces of this individually, either in the agent chat or with routines or with whatever,

48:27 what you’re trying to do is you’re trying to work that process out, right? You’re trying to go where’s my this thing first principles and you’re trying to think about the system, right? So the routine, the agent chat, those early inputs, you’re like at the beginning of this loop. You’re just trying to like figure out what do I want from the AI? What data sources do I want to plug in here? What context do I want to give it? And you just want to get a result at first, right? And that first result is like, is this good or bad? That’s where you come in and you’re like, oh, this is useful for me or this is totally trash.

48:59 And if it’s useful for you, you start thinking about what pieces of this are useful and what can I modify on the front of this system to make this output a little bit more useful for me. And as you start defining that loop and what becomes useful for you, start peeling those things that become useful and you put those into a dashboard. ask the agent to like turn this into a dashboard for me or turn this into a database that stores these for me. So you can have historical context of your uh routines or your output or whatever it is that

49:29 you’re getting from that. And that becomes the iterative cycle that you want where again you chat with the agent, you tinker with some of the early stuff, you figure out what delivers the good results and you start feeding that back in. And that really is what creates something that like over the course of a month you will get to a place where like now I’ve automated hours worth of my time versus creating some big system where you’re like look at all these inputs and look at all this data and look at all these cool reports I’m getting but it’s not actually functional right like there’s no takeaway there’s no like outcome that you can measure from that system.

49:59 >> I think it’s also like a good um example of how to leverage AI to get more of out of what you’re already doing. So what do we what do I mean by that? Well, we have to do this show. we don’t have to do. We enjoy doing the show. We happen to do the show uh a couple times a week. How do we get more out of it? We leverage AI so that all the transcripts, all the feature requests go in here and then we can provide that to product managers and they love this type of stuff. You know, otherwise they’re like sitting down one-on-one with different people and to try to get this information. So, we’re get adding leverage to their time and

50:29 we’re adding value to them. So, same amount of work that we have to do. We have this tool and like we increase the overall value. So, that’s leverage, right? leverages you same input more output and so I think this is a really good a good example of that and leveraging AI for that. >> Yeah. Yeah. And again this is where you can make this is where people are separating themselves. Manny you talked about like the 10x or 100x or whatever >> 100x dude 100x where we’re going. This is this is where you do like your ability to um anybody can connect their

51:02 Google drive and their notion and their uh you know meeting assistant and their whatever and just say take all of this and put it into some kind of thing. Your ability to then refine that into something useful, right? Your ability to just say, “Hey, we have this system now. How do we turn it into something that like does this thing and is useful on its own?” That’s where you’re going to make your like that’s where you become the 100x person, right? Like that’s where you become the the the the unfireable person inside of your org.

51:32 This is where your agency becomes like you know next level where clients just start to say like hey you know I need this person thinking for me because when they think for me they amplify the output of everything we do right and like that’s where you know going back to uh the question of like how do you turn this into a system I think like early iteration chatting with the agent creating these types of routines is kind of like how I sketch if I think about it in the art context I’m sketching I’m ideulating I’m running through things really quickly when it’s time to get the canvas out and break the paints out.

52:03 I’ve kind of already done my iteration and I’m ready to do them like build something that’s going to like last for a while that we want to put on the wall and maybe other people want to look at. So, I’m thinking about this as a process where like I’m using the AI to kind of sketch, run through ideas, pull things together really quickly and then as I refine that, I’m working towards the masterpiece, towards the thing I want to share with people or the team or like put into the production. So there’s this kind of like arc of, you know, how I start thinking about and when I start uh engaging the agent.

52:35 Um, okay. Any other thoughts or should we dump this next? Uh, >> I think we keep there’s a couple other like smaller questions that we’re going to get to. Um, but >> let me dump this in and then we can just go into whatever. Yeah, we can we can um handle all those as as we go through. But let Okay, so here’s what we have. Uh, we have Aura’s S1 filing here. So, there’s lots of data. There’s a bunch of spreadsheets that are kind of inside of here. And again, I want you to start thinking about like

53:06 you you can just kind of throw anything at the agent, right? So, here I’m going to just take this data set and show me or’s revenue trend over the last seven quarters. I wanted to chart how paid members have grown quarter by quarter. And I wanted to plot membership revenue and paid membership on the same chart. And I’m just going to dump the CSVs in there and let’s see what it does. And I’ve got a bunch of other prompts that we’re going to feed this thing. So, let me just look at my different prompts here. So, uh we’re just looking at some gross stuff. We’re going to ask about revenue mix and profitability and cash flow here in a second. So, um and then,

53:38 uh to to the point, we’re going to maybe follow on and maybe we’re going to turn it into either an investor presentation or a simple dashboard. So, um from here on out, we’re just going to totally experiment and tinker with this data set. um it’s going to do its thing. So to uh was it Nadia or Noni? I’m sorry. I don’t remember who asked about how to like take this to the next level. Hopefully between now and the end of the episode, we will take something from this data set uh and just start tinkering with it. Maybe we’ll get some charts here in the agent chat up front

54:09 and then maybe we’ll turn that into some kind of dashboard or some kind of thing. And so you’ll see the whole thought process from front to back here. Um and at this point, let’s see where are we at? It’s about an hour into the stream, so I think we’re just in free flow mode here. Uh, feel free to drop comments, thoughts, anything you might have, questions, uh, concerns in the chat. And, uh, Manny and I will just kind of answer those as the agent builds in the background and we’ll just kind of riff uh, and wax poetic on anything else that comes up. >> I’m I’m always surprised how closely

54:40 some of our users track all of our stuff. Um, here’s the comments here. Using the new add a collab on the demo. Yes. So, as we announced last week, we recently acquired a company called ADA. And what they’re bringing to the table is what you’re going to see in a bit is this ability to put together some beautiful inline charts for your data analytics. So, I won’t chat too much about it, Ry. I don’t want to like I want to see uh see what she do. >> Yeah. See what she does and show people that. But, I am really excited about this. Um, as some of you know, I have a

55:13 data science, data analytics background. And I remember having to learn tools like PowerBI, um, and Tableau and Python and R. And one of the primary reasons was because I wanted to put together these beautiful custom charts for all the stuff that I was doing. And so, you know, having demoed this for like the team demoed this for me last week, I was like blown away like, oh my gosh, like that should take me forever to chart and to create. And so really pumped to be showing uh what you’re going to get with this new collab.

55:43 >> Yeah. And I again like um creating slides, creating quick little visualizations, creating these little like dynamic things. And maybe this is the moment to um share this this deeper. Uh I hope he doesn’t watch this. He’s going to get mad at [laughter] me if I share this. By the way, >> don’t share his real name, dude. >> I’m not gonna No, no. He he’ll get like super mad at me. Um okay. But um and I agree with you, Manny. It’s wild like how in tune a lot of the audience is with like [laughter]

56:13 >> okay uh before the example here we are like it’s already starting to drop charts for me and again I want you to see you know like the chat here is happening and then assets it’s starting to create these assets on the side here um okay let’s see so it created a bunch of charts but it looks like it’s uh >> nah you gonna stay on free for a minute long yeah let’s just keep let’s just keep it in check here in free mode and see what it does then we might pop that up in a sec but I want to like push I want to push that a a little bit. Okay, so quarterly revenues. So, okay, before we get to the story, I’m going to keep teasing the story. Okay, quarterly

56:44 revenues peaked in Q4. Okay, here we go. So, again, I don’t need like some big dashboard for this. I don’t need to like um you know, let’s see. Oh, we we can expand this if I want to open this chart uh in a in a view here. Okay, we can go back. So, again, this is not just a chart in the chat. We have like a whole interactive little thing. I can even just kind of like grab and pull and play with this thing. This is beautiful to have in line, you know, >> in the chat to not have to create software, to not have to render or do anything and just have this stuff here.

57:15 So, okay, paid member increase every quarter. Oh, Aura is popping off. Look at this. Okay. Uh, nice membership revenue and paid membership both more than tripled. So, again, here are some nice simple charts. All right. From Q4, paid membership grew. Okay. The CSV doesn’t specify financial units, so treat it as revenue as USD. Okay, that’s great. All right, so yeah, there we go. And now we have these charts. If I want, I could download this chart, right? Here’s the downloaded chart.

57:46 Where did that go? There you go. You know, so you can pull this out and now I have a chart I can share. Let’s say I wanted to throw these in a slide deck. You know, let’s dive a little deeper before we start throwing anything into a slide deck. Actually, uh what happened to my >> Where’d it go? Where did you go? There it is. Okay. So, let’s uh let’s see what else. Let’s see. Revenue mix profit. Let’s just throw all

58:17 these in here and see what it does. I’m just going to throw them all in at once. There’s no there’s no need to like do these uh slowly one at a time. Let’s just ship it. Uh, go back to where was I? Nope. I was on this side. Sorry. Did I? Oh, here it is. Okay. All right. And I’m going to do this like All right. Let’s see some more charts. [laughter] Let’s not even get too Yeah, just let’s

58:47 see. Let’s see some more charts. So, we do, you know, you do need to think about your prompts, but we already thought these through. So, we can just send them and let’s see what it does. Let’s see how it handles this. Again, we’re in free mode here. Uh, and see how deeply this thing will go with this before um we bump it up. And then what I’m thinking, Manny, is maybe we take these and turn it into a little slide deck and maybe, you know, like an investor, you know, like a >> end of quarter investor presentation um to share with folks. And maybe what else

59:19 could we do with this deck? What else could we do with this data here? Like >> the other cool thing, two things to call out. Um one is Rhymark dropped in a zip of um different data sets. Um Replet agent can also research that for you. So, if you drop in like, hey, give me like reports and chats on Nvidia or or Meta or Google, it’ll go and grab those um resources um the actual financial documents and pull it into the chat and and and render the charts. The other cool thing you can do here is you can

59:50 customize the styling as well. So, if you want to say like, hey, like make this chart in the style of Nvidia or in meta user brand colors, that’s all you have to say. and it will create those nice stylistic charts for you. If you have like a certain way of doing that internally, um it will do that for you as well. And um there are some other like really radical charts that you can do in here as well like uh sanki charts. I don’t know if anybody else give me a shout out if you know what’s a sanki chart here. Waterfall charts and curious

1:00:21 has anybody had to previous to AI manually put together sanki charts? Uh if so, I want to hear your horror stories. But these charts, they’re beautiful. They’re great. Uh, but they were just like so hard to put together. You actually needed to learn R and programming uh and Python to actually do them. Now they’re just one prompt away. It’s so amazing. Um, so like a lot of beautiful and rich charts that you can you can build in here as well. >> Yeah. And I think that’s a good thing to push it further. And I think this is one of the things to point out as you’re

1:00:52 dealing with this is that you can just kind of again push the agent further. So like right here we’re asking it to just compare um the data that we sent in. But I liked your idea of like, you know, maybe pulling in some competitors. So, uh, show me how this compares to market competitors. Research the web and give me a benchmark for how this ranks related to other products on the market and other industry leaders. Seven course.

1:01:22 Okay, so it’s still doing its thing. I’m going to wait until it uh delivers this next one before we send this prompt because I think this is going to send it in kind of a different uh direction than what it’s doing here and I don’t want to interrupt it fully. Um um there’s a really good comment here from PayPal. He built a tool for a company that to get Excel sheets into a new form. Save them 20 minutes daily for admin. I use AI. I mean I think there’s like a still like we’re, you know, we keep saying this, but we’re all pretty early to this. There’s a lot of companies out there that are still doing stuff in a traditional format. there

1:01:52 oftent times it’s like in Excel it’s like all like like sometimes it’s paper and pen so if you can come in and automate some of that I like there’s like a lot of opportunity there so thanks for mentioning that PayPal yeah and this is where I think again like even if you do have to present a spreadsheet even if you do have to take a Google doc even if you have to take something that like already exists in the world you can use replet to make it better you know like you can use replet to clean up that visualization you can you know uh think about how to present

1:02:23 this in a more interesting way in a more dynamic way. Uh sometimes a spreadsheet can be like confusing. So if you have some complicated um if you have some complicated spreadsheet then you can um you know like again bring that into some kind of UI or you can bring that into some uh place where you can modify instead of manually entering the new digits create a slider that lets you see like what happens if we you know uh scale revenue like this or if we change

1:02:53 this input. So again, instead of a static sheet where someone might have a hard time understanding each of the inputs and how they relate to each other, you can take that spreadsheet and have the agent turning it into this little beautiful dashboard that still uses a spreadsheet as the source of data, not some database, and then give and expose those controls to the front or to anybody that’s using this thing. >> Make it interactive. Oh my gosh, what an opportunity there. Because you, you know, back to my data analyst days, you would present like a couple of suggestions and like here’s scenario A, B, and C, but then you always got these

1:03:24 follow-up questions like, Manny, what if we change this conversion rate or what if the growth rate is this amount or what if this happens to this other variable? And then you’d have to scramble like, uh, you know, we think blah blah. If you have an interactive dashboard, you can say like, oh, that’s a good question. Let me go here. Let me change this variable. Boom, in real time, here’s the new updated estimate. You’re just going to wow so many people in that room. Yeah. Yeah. And I think again like the um the ability to just push it, you know, one step further or to be there, you know, a half hour before the meeting

1:03:55 and trying to refine, you know, what the next little thing could be. Um yeah, that that that’s [clears throat] where this gets interesting is like the speed at which you can just kind of jump into the mix. Uh okay, so let’s see. Let’s close that out. source totals reconcile hardware plus membership. Okay, so they’re doing its thing. Okay, so I and if you remember, I kind of dumped a lot here, right? Like so it’s it’s uh we asked it for revenue mix, we asked it for profitability. There we can see it’s

1:04:26 adding some other charts here. So in real time, you can see that’s kind of updating. U let’s go see what we have here. Uh okay, so membership rows. Okay, great. And look how again look how these charts. So shout out to ADA and how this whole works as I hover here >> how interactive that is sales and marketing le operations. >> Can I say one more thing? Um the the way that this is rendering there uh like maybe two years ago there were external services whole businesses around this uh

1:04:56 I forget the name of the company but it’s like a a German based company if somebody in the chat knows but essentially it was used by data focused publications like 538 new uh the uh New York Times um the shot or I forget what it’s called but these publications that were doing data heavy reporting they would send this to an external service to route these types of charts. And I remember the looking at the subscription rar and I was trying to get it. It was like $3,000 for the seat for the year. And I was like, whoa, what

1:05:26 the heck? Um, and so now to have this here, like this is exactly the same thing I was like I was expecting to get from that service. Um, to have this here in real time, it just boggles the mind, dude. >> Yeah. U Yeah, right. Exactly. I mean to get this type of thing you would have typically had to build let’s say even to get this thing in a in a in an AI or here in in replet before this you would have had to like build the whole UI and like feed this into a system and have some kind of rendering engine and have some kind of like recharts or some line

1:05:57 you know some kind of charting tool and so to be able to do this uh you know right in the chat um with this level of flexibility to pull these out um what would be interesting is to tell it to just throw these all you know again in a slide deck I’m going to go ahead and nudge this thing uh steer is the proper term because it’s just going to run on. So, it’s looking for other things. Again, if you remember, we we fed this thing a ton of prompts and so it’s just creating all these different charts for us. But I want to see, hey, can it look up market research? So, let’s just see

1:06:27 what it does if I uh throw in an extra prompt here. What you’re also watching the ability to do there is what they call steering the agent. So instead of stopping the agent, instead of uh starting a brand new task here, I’ve just kind of given it additional instructions. Um and this is uh you know a fairly new feature where you can uh again steer rather than restart or stop the agent uh midtask. Look at this. Acred liabilities help lift operating cash. Okay, great.

1:06:58 Scroll to latest. So it’s still creating these charts. >> Nice little waterfall chart. Those were >> those were so hard to do. Like we have to use used to have to do those in Excel. It was like a complicated mess to do those waterfall charts. >> Yeah. Okay. So, all right. So, it it has now completed the uh you know these batch prompts that we threw in here. Uh all these different This just blows me up. >> Oh, wow dude. >> Um okay, so then what we asked it was Manny suggested, hey, how does this

1:07:30 compare to the market? uh you know like what other research cuz right now we were just kind of working off of like their internal filing. So this is like a fixed data set. So this is like a data set that maybe um you would pull at your enterprise or you know like something you want to work internally. But here we’re going to ask it how does this compare to the market competitors? So I want you to research the web. Now it’s going to go out and maybe look at other public filings for companies that are in this same space. Um, and we’ll see what it comes back with charts on that. Um,

1:08:04 and I see we lost our comments and talking about automations, huh? [laughter] >> That’s the thing. You got to maintain these. That’s the thing we should probably have another show on, right? Is like, okay, great. You launched your app, now you got to maintain it, [laughter] >> right? >> But I can bring lower thirds. I did find the names of those companies that I so one was flourish and I dropped in the chat and also the other one was data rapper as well. Uh data rapper is the one that I like. So if you look at data rapper look at their charts and then look at the stuff you’re getting here in the chat it’s like oh my gosh and it’s

1:08:35 so much faster crisper and if I look at the pricing here like uh yeah got like 40 bucks a month now. Oh they have like monthly pricing now. They didn’t have that before. Um but yeah just yeah those examples there. >> Yeah let’s see I’ll try looking up the Aura CC filing maybe the Whoop financials to clarify their market position. So it’s identified Whoop as a competitor. Uh additionally I should search for IDC’s market forecast and reputable sources like counterpoints. Okay great.

1:09:05 >> So you can kind of come through and see and this is where it’s you know like this is where it’s doing its thing right locating original market share study. Okay. Search 10 websites clear as comparable so far as paid membership or reports. Okay. So it’s looking at like now data and what uh like what it might compare to to bring in. So again this is one of those things that you could just have running in the background. Um you know we’ll look at what this costs to run here at the end

1:09:36 of this thing. Uh we are in power mode uh at this point. Um but I just you know like going back to the the concept of knowledge work here right like it is now your job to think about not just hey we have these AI tools at our disposal but you now have a mandate to think about how do I effectively apply these tools in a way that like amplifies the dollar I spend on them and the output that we expect as a group using them right like we’re no

1:10:07 longer in this like Kenned work phase like we’re beyond that right like we know these tools work we know you can build almost anything uh we know you can kind of like again approach and ask for almost anything inside these chats the question and the onus now becomes on you to shape what it is you ask for the data you provided as context and then really that human iteration inside of that to get to the desired outcome right Like

1:10:38 that’s >> where I think you’re going to make your money over the next handful of years, right? Because if everyone has access to these tools and everyone can build, then what’s going to separate you is your ability to use the tools in a in a little bit better way, a little bit more, you know, efficient way or to again become one of those 100x rather than 2x, right? The other thing I would add to that is that um using these tools is an

1:11:08 opportunity to refine your thinking and to improve your communication because if I think back like okay as a data analyst as a data scientist what was like the highest value thing that I did is I came up with a recommendation I came into a meeting I presented it clearly people got it and people trusted me in that recommendation. they trusted like the work that I did and how I explained it and my judgment that that was the right recommendation. And so I think that’s kind of what you have to optimize for. You have to use all these tools to go

1:11:38 through all the different variations. Do all your research. Don’t just rely, you know, you got to like double check agents as well. um do all the visualizations, go through all the different scenarios and then think about like what is that twominut like because you know we say like hey like you can prepare like a 20 deck presentation but really what people care about is the last two slides your recommendation and why you got there. So like really refining your thinking around there and then being prepared to support that when you get push back on that recommendation. So like as I mentioned earlier like Manny have you thought about this? What about this? do this

1:12:09 like yes yes yes yes here’s why y why why oh oh like that’s a good question let me pull up my tool here’s my you know here’s how I would change this boom like that’s credibility that’s like going to help me get my point across so I think it’s like can you like more refinely like better articulate that recommendation and can people trust you on that yeah and I think I mean this is the same type of thing as anything else right like if you like let’s say you’re a writer and you know like you could write something in a hurry.

1:12:41 You could hit the deadline. But if you write something and then you let it rest, let’s say overnight and you read it the next day and you give it another look, the output or the quality of that initial piece is going to go so much higher than if you had just written it immediately and fired the bullet. Right? And this is where the AI kind of thing is, you know, like useful in the same context where like again you should use the AI, you should use these tools, you should use these things to kind of help you think, help you explore, help you sketch, help you kind of iterate and

1:13:12 then you take those inputs and feed it into the next phase of the thing. And even if you use the AI in that next phase, you are then like kind of thinking through how to maximize the quality. You know what good is, right? like uh and so that becomes your job as a knowledge worker to amplify this rather than just you know taking the output and assuming you can just forward that on and firing slop cannons at you know the rest of your team. >> No, somebody came up with a new term the founder of uh or and CEO of uh Shopify. He said um slop grenades.

1:13:44 >> Slop grenades. >> Yeah, people are lobbing slop grenades. I think I mean it’s a good topic rammer is like okay like I think my feeling is like who’s responsible for the AI slop and the slop grenades that’s still you like if you’re using AI you still have to know how to use it in the right way you just can’t give it a prompt and like take exactly what it outputs you and then send it over the line like you have to like make double check it you have to refine it that’s oftentimes what I’m doing I’m like leveraging all these AI tools but I’m I’m look I’m giving it a final pass on everything I’m making sure it’s nice and tight I’m doing the final

1:14:14 pass refining it and then I’m sharing sharing that with people. I think that if I just like grab an output from an AI service and then send it over to you, Rhymar, or to someone internally, like that’s not good. That’s not good practice because then it typically it’s going to be like overly verbose. It’s not it’s going to give too much detail. It’s going to like not be like like know like level and it might be you might have some errors in there. So, I think ultimately like you are still responsible. You’re like the editor now. I guess like like like the writer sends the thing to the editor and the editor has to prove everything. Make sure that it’s like everything’s fact checked and

1:14:46 ready to go for publication. Like that’s now you with everything that’s coming from your AI tools. >> Yeah. Yeah. 100%. Uh limiting AI slop. Yeah. Right. [laughter] I think it’s our responsibility each of us to, you know, remove one piece of AI slop from our world every day. >> Yeah. Yeah. >> Oh, look at this. >> A slop. Hey, look at this. Look at this. >> This is called a tree map. Um, this is really nice. Um, we used to use this for market cap. So, like if you take the S&P 500 and you break it out by market cap

1:15:17 by company, you see how much they drive the action of the overall marketplace. And so, like you have this now here. Really neat. This is cool. >> Yeah. And again, like uh right, let’s see what how responsive it is. >> Can Can you do this? Can you >> responsive? Like I don’t know if you’ve ever done one of these and tried to like ask it to reformat the way the blocks lay out. Like I’ve I’ve tried to do these types of charts in the agent before and um you just it’s not this nice. >> No. Yeah, it’s not this nice. Yeah. Might might have to stay in chat mode

1:15:48 for longer. [laughter] Um can you ask to do a sanki chart of um like cash flow from like revenue to expenses and see what we get for that because I love I love people to see a sanki example. Sanky chart example. >> See uh let’s see our rank for smart. So now it’s going to create like a little overview here. Let’s see. Uh whoop disclosed more than two and a half million members. Okay. Garmin’s entire fitness segment reported. So again, you know, you you got to go

1:16:18 through and fix this. I wouldn’t go and try to raise money off the the full output here without, you know, taking a pass like you said, Manny as being the editor. Um but as a starting point and we’re obviously using the data from their you know uh or as results here that we uploaded directly I’d be interested to yeah see and a buck 75 to do this research worked for 14 minutes this thing worked in the background for 15 minutes uh created all these different charts here so what’ you say what’s the

1:16:48 name of the chart >> I just dropped it in the chat um I I think I really like this example you’re showing because you know leverage there’s a couple different ways where I see power users leveraging Replet in ways that others aren’t. Number one is how you mentioned it here as using Replet agent as your research assistant. So agent has access to the internet. It’s going to go find stuff. It has context from your from your conversation. So it knows what’s relevant to you. So let it go and rip and read all the articles and then bring it back to you. You have to double check

1:17:19 some of the sources obviously. And then the other way is leveraging replet agent as your business al um agent or excuse me or analyst or strategic advisor. So now that you have all the context here say like okay based on all of this detail do you think that Aura is a buy or a sell? Tell me why. What are the recommendations? What are some of the risk? Should I buy this? Here’s my here’s the other names in my portfolio. Would you recommend me adding this? Is this properly is this properly diversified after I add it? how much

1:17:49 should I add to it? So, you can also use agent to give you those strategic advice. It’s like in one place, dude. You’re like you’re like I think you’re like what you’re you’re like you have like five professionals here in one place from the analysts to the strategic advisor. Um uh it’s just like it’s just like all in one here in in this session. >> This is a uh how do I copy this? Horror decided to delay its IPO this morning. Any chance we can ask the agent to showcase why?

1:18:20 >> Let’s do it real time. That’s what I’m saying, dude. It’s your real time research assistant, dude. >> Okay, let’s see. Uh, all right. So, Aura delayed their IPO this morning. Oh, snappy dude. Look at there’s your chart. >> Oh, look at that, dude. Okay, so what’s cool about this is that you can map like you have revenue coming in. Oh, uh, you can drop that prompt in there. Then we’ll come back. No, go ahead. >> Uh, all right. So, like here, like you

1:18:52 have all the revenue that’s coming in. So, think about like your paycheck that comes in. Where does it go? Okay. So, co cost of revenue. This is what it cost you to run the business. So, this is like hardware for Aura, data analytics, whatever. And then here’s the overall gross profit. Here’s a chunk of it. Okay. Where does the gross profit go? Well, then it goes to overhead, operating expenses, and then we have operating profit. Okay. So, then where does the operating expenses goes? Well, it goes to sales and marketing. goes to research um um general administration. Okay, cool. And then where does the operating profit go? Usually operating profit either goes back into the

1:19:23 business or goes back uh to uh buy back shares or to dividends. And so you can clearly see like all the operating profit that comes out here and where that goes. And so this is a super powerful chart when you’re trying to understand the overall healthiness of a business because sometimes a business might have a high topline revenue, total revenue that we see here, but they might have a narrow profit margin, but like all of that margin is going to back into the business or in dividends, but that might be a good business. Sometimes people have low topline revenue, low revenue, but it’s going back uh into the

1:19:54 business as well. So like there’s different things you want to look at to to to gauge whether or not this is a good investment for you. And we used to do this all the time all the time and they were like so tedious to do. Again, I this was like a custom um function that I had in Python and it wasn’t interactive. I couldn’t do this here. I couldn’t customize. It’s like a static image that I had. I had to copy paste that put in the presentation and then I had to do that every time every quarter for different companies and then there there are over 500 companies in SP500.

1:20:25 That gets really painful really fast. So like you could probably batch this stuff up and like just have these cranking these out. Um, it’s just it’s just amazing. It’s crazy. >> Yeah. Yeah. This is great. I didn’t even uh know this was called a Sanki chart. I I knew of this chart, but I didn’t know what it was called. Okay. So, uh while you were sharing that uh insight, I prepared another prompt here. So, Aura delayed their IPO this morning. Take a look at recent news. Analyze the data provided and let’s come up with a percentagebased ranking with reasons for

1:20:55 why this might have happened. Use market conditions, recent financial news, and anything else related to why this decision might have been made. Do not just look at official talking points. Dig behind the scenes and look for reasons which might not be obvious on the surface. Let’s just see what we get. >> Go deeper. >> Go deep. >> Yeah. And so this is again one of those like push the agent, right? Like like Manny was saying, like you get the initial output, you think maybe it’s good, maybe it’s not. Uh you know, like you put your slot detector on it [laughter] and figure out whether this is something that you would be proud to share as your thing, right? This is the

1:21:25 this is the benchmark that I kind of look at is like would I be proud to put this out as if Rhymar had like forget about AI. Just assume nobody knows how Rhyar got to this output. If I put this thing out in the world, is this something Rhyar’s proud of having come from him with whatever standards I have for the work that I do, which are incredibly high for like the finish output of like the thing I try to create in the world, right? And so that should always be your your buffer here. Just because you’re using AI doesn’t mean that you should lower that standard,

1:21:56 right? Or that you should lower your quality requirements for what you put out into the world. And I think again, that’s the that’s the secret sauce here is that you need to kind of make sure that you’re the the the definer of that thing. Okay. Research notes. Uh all right. Let’s see. Plan question. What evidence best explained? Scope, current reports, audience. Yo, this is creating like a whole type of situation too that I’ve not [laughter] I’ve not seen this as a as a

1:22:26 report, you know. >> This is new. Yeah, it sits as an artifact as an asset. So, you go and like reference it, check things off. >> Yeah. And it’s a kind of like a doc asset here. Uh look, it’s got some markdown so you can see how it’s organized. You can see the view. Uh so, again, it’s going to kind of just keep doing its little thing and it’s going to update this as it goes. That’s uh it’s looking for news and research right now. Um, let’s see. Two commands, 36 red, deep research, loaded the skill, load the skill. This is where I go back to

1:22:57 again like with a lot of these other platforms, you got to have a research agent and you got to go define that and you got to go create a skills.mmd file or whatever the blahy dah to have your research assistant like do what it needs. And then you need to come over here and have another like web search agent that you build. And like you know like there’s this idea that like you want to manage a bunch of like you don’t want to man manage a bunch of agents. I don’t want to manage the personality of my agents. I don’t want to like write skill files for every little piece of thing I want to do. I want to just come and ask the agent, right? And so that’s

1:23:29 one of the things that you get here. It’s like it’s all these skills are baked in like a lot of these things are already happening. You know like again one of the things with Replet is that you don’t have to think about all these other things. And um maybe this is the perfect time to like uh actually deliver on this story here. Okay, super uh bull signal for the industry and kind of what’s happening. A friend of mine reaches out this weekend and sends me this like long crazy message and he’s like, “Oh my god, I’ve been building blah blah whatever. I spent $1,000 on uh codeex and this that the other and I

1:24:00 built it’s crazy. I’m going to make a million dollars with this thing.” And so immediately my mind is like, “Okay, we have reached peak market saturation. Auntie is on Facebook, right?” Like we we kind of knew Facebook was dead when like all the aunties and moms got on Facebook, right? Like that was kind of the moment where everyone knew and so it’s like but it also meant like hey social media is now like in the world right on teaser on social media. So this to me was kind of that same signal in the like vibe code agentic engineering world where like my friend who is a designer but has absolutely zero technical knowledge is now building

1:24:33 tools. And I was like okay whatever but he knows I’m over here you know doing the replet stuff. and he’s like, “All right, I need your help.” Because he built this beautiful thing and I’m looking at this product. He’s showing me a demo and I’m like, “Dude, this is super cool.” But it’s like a on local host. It’s just on his machine, >> right? >> And now he’s trying to figure out how do I get this thing into the world? How do I deploy this? How do I put a database behind it? How do I get this thing off of my device and into somebody else’s hands? And I’m looking at him and I’m

1:25:04 like, “Bro, you have created a monster here that like is going to be very difficult to do this thing you want to do without again like authentication. How do you manage this?” Like sure like now you can download it, but now everyone has access to this thing you’ve created. You have no way of gatekeeping or charging or putting it behind a place that you can then update it and people can kind of regularly grab that thing, right? So like he just has this thing and it’s a cool thing. It’s a nice thing, but now he has this whole problem of like how do you take it to the next

1:25:35 step? How do you get out of this prototype phase or this like I can use the tool phase? And that’s where again replet just clears the path for you all the way from prototype to production to click the one button, you deploy, pick your type of, you know, deployment and it just happens for you. Um and so yeah, again while this is a sign in my mind that you know like the the world is opening up and people who are not typically building things are now looking at the things they can build there is still this gap of understanding

1:26:06 that like from a technical standpoint building and then getting that out into production is two different things right Manny we talk about this like he has distribution too like the guy has a YouTube channel he will get this thing out into the world but right now his concern is how do people log in how do they connect to it? How do they how does he manage the user’s access to this thing with licensing and all this other things that like you know how do you get it off your computer? Um, so again like the platform matters, how you approach this stuff matters. And if you’re not a

1:26:38 super technical person, if you don’t know about how to deploy things and how to set up your own servers and how to like manage your own infrastructure, your security, like all these other things, you should just be using something like Replet that just handles that for you, right? >> Um, spot on, right? Well, I’m I’m actually glad to hear that cuz to your point, it’s telling you that vibe coding and building with AI is going mainstream. And that’s great for just a lot of reasons. I think the sooner that people get some experience with building with AI, like it just kind of clarifies

1:27:10 to them like really the potential and how beneficial AI can be and it helps counter some of that AI dumerism that’s just kind of everywhere um right now. Um but yeah, I think this is why I think Replet is the the best place to start when you’re early on in that journey, not only for the publishing side. Um um Rhymar, but also on the security side. I’ve mentioned this story a couple of times. I was at a an event and someone like this was like a like a venture capitalist person. He vibe coded his first app and he and then he managed to

1:27:40 deploy it, but then he it got hacked like two days later. Um and I, you know, I went up to him like, “Dude, you didn’t build it in Replet, did you?” He’s like, “No, I didn’t. like I did something else like more look like okay yeah well those tools are meant for professionals that know how to lock down their their thing like replet assumes that you’re coming to this for the first time with no coding experience so that means the publishing is going to be easy the security you’re it’s going to be on guard rails there’s going to be also additional features for you to soup up your security but it’s all things like databases authentications like it’s one thing to have a nice front end but then

1:28:11 you got to publish that you got to get authentication for that you know if you’re doing it elsewhere you got to integrate clerk that is such a pain in in the butt. You’ve got to stand up your own database. That is such a pain in the butt. You also got to manage git and history. Like good luck going doing a roll back and like telling your friend like oh yeah you you have to go back to a previous branch and stuff like that. Like he’s going to be like what are you talking about dude? So you know for all those reasons I say like it’s great to start with replet. I think you know if you want to go elsewhere after that but I think like a good jumping off point for everybody is really replet because it is designed for that new type of

1:28:42 builder. >> Yeah. Yeah. And I again this is one of those things where you don’t know what you don’t know. right? And you start learning and you start thinking, you start building and you like get to something. Um, and this is where again like trusting the platform, knowing the platform can get you to where you need to go and not having to think about like where am I going to hit this bottleneck, right? With a lot of these vibe coding platforms, you’re going to hit a ceiling, right? like you’re going to like Manny was saying uh whether that’s

1:29:14 yeah maybe you can wire up your you know database but how do you split production versus development and then what happens if you want to iterate on that development deployment but not touch the actual development data or the production data well when you create or spin up a parallel task in replet you duplicate the development database and so you can tinker play with the schema mess with the data the test data you have in your development database with zero implications on what happens in production or even in the original

1:29:45 development environment. You have a completely segmented forked branch, however you want to call it environment that replicates that database for you and allows you to do what you need to do with no fear of deleting production data or uh breaking your schema. And then when you want to merge that back in, you don’t have to worry about merge conflicts or like how this thing syncs back in. You just click, hey, let’s go, you know, hit the green button and it brings it back in and like those the merging and those conflicts

1:30:16 and dealing with these types of things. Again, unless you have like a really high level of technical understanding, you probably don’t want to be like facing these types of things or these questions. And this is where, you know, uh, Replet makes its its its name. >> One more thing, I I see the ready. I’m ra I’m rich and like look at this cuz I saw some interesting insights. Go ahead. >> One little thing I just have to add but not only like why I started with Replet and why continue to recommend Replet is because it’s pretty straightforward to get started with Replet. We assume you don’t have a technical background but

1:30:47 also we’re also assuming that you might hit a rocket and you might have to scale really really fast. And so we’re seeing other folks start elsewhere and like they get traction and they’re like I had to jump off of those like maybe like easier to use or like more like hyped up platforms. I had to come back to Replet because I hit 100K users and everything started breaking down or like I needed more security or whatever. So, it’s kind of the best of both worlds. Like pretty straightforward to get started, but then when you do hit that rocket ship, which we’re all like betting on all of you to do, then you can like bet on Replet to

1:31:19 help you scale things out. >> Yeah. Can I give a shout out to my my buddy Zack Buazia here? >> Yeah. Is he here? >> From the old uh Web Flow days. Yeah. >> What’s up, Zack? How’s it going? Welcome. likes on the video. He says, “Let’s get let’s get some likes on the video.” We never beg, but come on. I I bet if you refresh there’s more, but uh yeah, let’s let’s let’s hype that up. >> Rhyma, you got like a whole Web Flow refugee group going on over here like on all of our streams like here coming from Web Flow [laughter] like helping to save my business with with Replet. Now,

1:31:50 >> what’s funny to me is that uh Zach and uh like he he made a comment earlier about the pineapple pizza stuff like there was a whole pineapple pizza fiasco in the Web Flow world, too. Oh, he run that over, too. >> That’s carried over. So, yeah, [laughter] >> I’m gonna escape it, dude. >> All right, let’s go back to screen share here. Let’s see what the agent has come up with on why Aura has postponed, not withdrawn their IPO. My read this looks more like a price and timing decision that evidence Aura’s business suddenly weakened. Okay, that’s great. Uh, but

1:32:22 that explanation is plausible, but the proposed deal structure gives investors another reason to scrutinize the price. And up here if we look uh market in uncertainly alone 73% of proposed shares were secondary sales. So you know uh people internally are selling the stock but you know maybe the market’s not buying. >> And I think what would be interesting to see is like um okay my a narrower API window. So here’s what it it gave us. We we told it for we said percentage based reasons, right? Give us reasons and then rank them on some kind of

1:32:53 percentagebased scale. uh the Fed raised rates. Okay, >> I think that’s probably, you know, like uh you know, we don’t need to get into financial speculation, but I think uh some of the same reason we’re seeing a lot of the AI uh regulation requests is because the market’s kind of getting tight. Uh people are, you know, getting a little more worried about what’s happening. Obviously, the Fed’s doing its thing. Interest rates are a little crazy. Inflation’s kind of climbing back up. Um so maybe that has something to do with it. Okay, that’s a 35% weight is a sign of that. Mhm. >> Valuation and seller heavy deal design.

1:33:25 Okay. Is this thing holders proposed selling 73% of the 50 million shares uh at 42 midpoint or expected uh 532 million in net company proceeds but earmarked 526. So they came in short of their expected earnings >> uh which might have triggered something right. So that might have said hey maybe these shares aren’t you know worth as much as we want. If we go out to the open market and they sell for less than that and the whole thing drops then you know that’s a bad look. So maybe they pull back on that. Okay, that’s interesting. >> Repeatable profitability. Okay. So, so

1:33:56 again, like this is just one of those things where like now you got to dig in. Now you got to go look. Um, but this was, you know, quite a bit of of work. It did here for a buck 64. It worked for 9 minutes. It did some research for us. Um, if I was like, you know, thinking through this or wanting to look at what this might be, uh, between this and our different assets here. So, if we look now, we’ve got all these different charts that we can look at and pull from. Let’s say we wanted to create a presentation. It’s there. And then we’ve also got these executive assessments with tables and data inside of here. Uh,

1:34:29 again, none of this is a piece of software. We could run this again. We could download these docs. We could now take these as a starting point for some kind of report we want to create. Here’s some sources if we want to investigate. So again, uh things that you typically might have gone to, uh another chat window or another LLM service, but with the benefits of Replet where you have your project context, you have the ability to like research. Now, we’re we’re using right now uh an external data set, but this could easily be just

1:34:59 chatting with a project you have internally and you’re chatting with your own stats, you’re chatting with your analytics, you’re chatting with, you know, your your database. So, right now we’re kind of like working with this public data set related to Aura, but this could be your own internal project. This could be any data set that you’re working with. >> Um, and again, you can just keep asking to push it further and further until you’re ready to turn one of these assets into a durable good, into some kind of uh piece of software, into some kind of dashboard, into some kind of uh maybe that’s what we’ll close with is like um

1:35:32 uh command mode is to off. Hang on. Um okay. Uh let’s take all our learnings from today and turn this into a investor presentation. >> Mhm. that covers a key financial metrics for uh folks who might want to invest

1:36:04 as well as uh concerns about market conditions and the pulled or delayed IPO. >> [snorts] >> Okay. And let’s just see what this do. Um, turn into investor presentation. So, >> [clears throat] >> oop, finanil. Uh, you know, [laughter] >> smart catch. No problem. >> Yeah. No, for sure. Um, >> okay. So, now it’s going to put it into an investor deck. Uh, we’ll see that

1:36:35 created as a new asset. This will probably turn it into I wonder if it’ll turn into a project now. Uh, because, you know, now it’ll start creating uh other um artifacts here. So again, um, you know, front to back, I think the the whole goal of this episode, if you if you’re still watching with us from the beginning, is just to go from this idea of like, hey, the knowledge world is evolving. Here’s how I can approach and work with these tools. Uh, all of the insights that you could generate from this will uh translate to any AI tool that I that you’re working with, like

1:37:06 thinking through your prompts, adding the right context. where you get the benefit again of working with Replet is having the project, the context, everything fully integrated and being able to go from let’s chat about this stuff all the way into like a full scale like let’s say we wanted to turn this into a an investor dashboard that just monitors this one company. We’re going to keep feeding it these new charts and we want to turn that into a durable good that I want to share with my investment team. And now we’re going to have like a running live dashboard that this thing becomes based on these learnings and all

1:37:37 these different data sources that maybe we find live data or maybe this is an existing project and we want just a a snapshot for folks who don’t really care what’s happening in let’s say in our community but they want to know what kind of activity is happening in in Discord. So maybe we give them just a you know just a little glimpse at that one little personalized thing. And so like these are the types of things you can start looking at and creating um >> specifically without the fear that you’re going to hit the ceiling, right? Like a lot of times when you’re building this stuff like there’s this idea again where my friend is facing today is like

1:38:07 hey I know I have a cool idea. How do I what do I do with this now? >> Right? Like you shouldn’t have to ask that question. The platform you use shouldn’t ever leave you in a moment where you’re like what do I do now? Right? And so as you look at like suggested features, uh, ambient intelligence in design mode, you know, like all these different things that Replet does is about putting you in an environment where you’re not going to hit that ceiling, right? Very low barrier to entry to get in. Anyone can start tinkering, but then the technical ceiling goes as high as you want it all the way to where you can bring in, you

1:38:37 know, like a full dev team and they can like start playing inside of this because it started as an IDE. So there’s no like, you know, you’ve got this whole band here where a lot of other tools fit nicely inside of a small band, but they don’t give you that full spectrum. So, um, anyway, that’s my my pitch for for the day. Uh, this is what knowledge work is evolving into. [laughter] And the better you get at using these tools and working with this type of thing is, you know, again, is going to amplify your potential in the next few years. There are a couple of questions

1:39:08 from Brett around how to validate the results, the analytics um results. Um couple suggestions there, Brett. Like number one, I think it’s helpful to have a sense of like the underlying data at least from like my like my um experience to have like a couple of like tests that you can run if you like if agent gives you an output like a summary like be able to like have like a unit test of sorts for agent uh question that you have the answer towards already or like a couple to make sure that agents pulling all the right contacts the right data sources and then also you saw in

1:39:40 the write up um agents also giving you sources also like check like double click on those every so often make sure like they’re good links to pull in the right things. Um, and so that’s kind of like just like like um my my [clears throat] suggestion. I’ll tell you that this uh even happened before the days of AI because we used to get like a data source and like you would assume that it’s a good data source but sometimes it was corrupted or you would like mix up your data sources like you you name them like differently like it was like this is Q2 versus Q1 and so I think this has just always been part of

1:40:11 the analytics workflow just validating your work making sure that you have the right pipelines going in. no different. Now, ultimately, you’re on the hook for that recommendation. So, you better make sure that all the underlying pipeline that goes into that recommendation is tight because again, like you’re on the hook for that recommendation that you give. >> Yeah. Yeah. And I think again like you’re on the hook for all of it, right? Like you are the like you are the end like the nobody’s going to sue an AI, right? >> Right. >> Nobody’s going to sue an AI. the AI is not going to get fired, you know, like

1:40:42 if you’re just firing slop cannons all day long and works like, “Oh, you know, what are you doing here, Rhymar?” Um, so, you know, like you become the, you know, the again like the the uh the taste maker, the the the thing that makes this worth looking at, right? Um, nobody needs extra paragraphs to read, you know? So there’s that whole meme of like take these bullet points and turn into a paragraph and then on the other side somebody’s like take this paragraph and turn it into bullets. And so like you know we’re kind of beyond this phase

1:41:13 in my mind where it’s just like let’s generate all we can generate and now it’s about like how do we filter down what we generate? How do we again work with these tools to like focus the result of what we want into something useful? And then how do you as a professional knowledge worker get better at like [snorts] that loop, you know? Uh there’s another question from earlier in the chat. Can this be uh set as a dashboard within the chat analysis or outside of replet? And I

1:41:44 think that’s what you prompted agent to do. Now I guess the question is uh you do all this analysis. This is knowledge work now. You’re doing a lot of time like thinking about the problem different angles looking at different charts. Now you have to summarize all of this and save it for yourself for posterity because two weeks from today you’re going to forget everything you did today or how you’re going to present this to someone. And so what Ramar prompted here is for a PDF here. So you can output as a PDF just oneoff report. Uh there’s another prompt going on for a dashboard. You can also do a slide deck.

1:42:14 So like if you want to share like a PDF of a slide deck, you can do all that or you can just have a report that gets written up in a nice markdown type of format as well to share with folks. any number of ways. I think the question is uh how do you want to present this information? How do your stakeholders expect this analysis to come their way and whatever that is? It’s just going to be an easy way to do it in replet. >> Yeah, let’s take a look through these slides here and see what we got. Uh ARA investor lens independent briefing. Let’s see. Supplied financial CSVs. Our statement SECS11A.

1:42:45 >> I think we should all we should say that this is not financial advice everybody. [laughter] This is just for exploratory reasons. So like please don’t like do take any bets on anything. Plus this IPO isn’t even live. So you can’t even unless you’re like doing like some VC stuff. Uh not financial advice. >> Well and it was a good uh call out that you said Manny that this is like uh you know uh delivered in a PDF. Uh we could have this create this as a um as an artifact in a project and then it would be more editable. We can you know ask it to iterate from here. Uh but you know

1:43:15 this is not the let’s say slides UI. if you ask the agent to create like a full interactive slide deck. So this is just a PDF output here. Um okay so let’s look at what we got. Uh act attractive growth unresolved entry point. The postponement should not be confused with the collapse in demand. The investor question is whether growth margin durability structure justify the proposed. Okay. So it’s kind of like laying out like hey we’ve got this scale. We’ve got some momentum but there’s a little bit of a risk here. Okay. That’s interesting. Revenue base has scaled sharply. Uh

1:43:45 trailing four quarters. Okay. quarterly revenue USD millions fiscal year ends. Okay. Revenue mix membership is growing. Hardware still dominates. So it’s showing what people are spending on membership versus the hardware. Okay. Members retention. So >> 85% retention numbers. Okay. Gross margin improved. Operating leverage did not. >> Mhm. [clears throat] Gross margin. Operating expenses. Oh, expenses went

1:44:16 way up. >> Go up. Yeah. >> Mhm. >> That’s like marketing, overhead, admin stuff. >> And then operating income went from $9 million to they lost $5 million. That that would be a reason to not go public with the Yeah. Cash generation is positive. Capital outflows were larger. So >> interesting. Yeah. So they’re they’re spending more than they’re they’re bringing in. Uh probably trying to grow and take advantage of this AI moment or whatever. But this is not a proof of liquidity crisis, but working capital moments, buybacks, and new debt deserve review alongside headline operating cash

1:44:47 flow. So again, this is just kind of laying out the case for why they might have pulled the the um the IPO. So um and I’m not a financial guy. I I have no idea, you know, like what the um you know, real reason is for this. So don’t sue me, Ora. [laughter] Uh okay. So yeah, growth is priced at roughly 11x trailing sales. Why buyers may pause June quarter revenue still comes from hardware with a quarterly operating loss. Okay, so again you know like uh great starting point way to just

1:45:18 dive in. Um you saw us if we went through uh the beginning of the chat here we started with just a uh data set. We asked it to kind of show trends in revenue. Uh that started creating charts for us. So we explored some of those charts and then we asked it to create more charts uh related to profitability and cash flow and the balance sheet. Uh so it walked through that and then created a bunch of other interactive charts here. And then what we asked it to do was to go through and show how it

1:45:50 compares to the market. Um and then somebody in the chat basically said, “Hey, Whoop, uh or not Whoop, um Ora delayed their IPO.” Uh and so then we asked it to kind of give us an an analysize uh go ahead and analysize why that might have happened. And so here we are kind of looking through that and the comparison. Uh we asked to create a Sani chart uh for revenue to cost visually. Um and that’s this chart here which I thought was uh super cool as well. And

1:46:21 yeah, so you’ve kind of just seen us explore this data set. If we open this up, we’ve got all our different assets here. So, we’ve got all these different charts that we can export, modify, ask for more details on. We’ve got a couple reports in here that are like executive summaries that give us uh percentage based rankings and we, you know, asked this all through the agent chat. Uh let’s see, investor briefing. It did create the slides as HTML here. So, uh we do have them as HTML and a PDF that you can just quickly download. Um so,

1:46:52 again, all of your assets are here. We did not stand up any project. We didn’t talk about deployments. There’s nothing here that needs to persist outside of this. Um, but it’s just a great example of how you can work with the chat to take a data set external, internal. Uh, this could have been a replet project that already exists and you’re asking about the the data inside the database. Uh, we just vetted some CSVs of these public um public performance from Aura. Um, let’s see what else did we do? Oh,

1:47:22 we we created this automated um >> routine. Yep. That kind of gives us a breakdown of what’s happening inside of Slack, Granola, my Google calendar, and notion. Um and so you’ve kind of watched us walk through quite a few capabilities of what the agent can do, uh how you can think through your prompts, where the future of knowledge work is headed. So, um I think we’ll leave it at that. If you have any questions, uh feel free to drop them. We’ll we’ll hang here for just another minute to see if anything else comes in. Uh if you’re not in

1:47:55 Discord, you should jump in tomorrow. We’re talking about pricing your Replet projects. So, how do you price the things you build in Replet? We’re going to talk about the different business models, uh positioning, uh the different pricing models themselves. Um and so I think that will be an exciting conversation and we’ll answer any questions as we follow along to this. We like to carry over uh the things we take on the Tuesday deep dive into office hours. So if anybody has questions about this and you want deeper insights, we will also set some time uh aside in the office hours to do that. But Manny uh

1:48:27 this was a deep dive to be sure folks are still hanging with us. They are nerding out on data analytics and charting and how to use the agent. Uh so hopefully that was useful for folks and um yeah, we’ll we’ll we’ll leave it at that. Any any closing thoughts from you, Manny? >> No, I really it was a great um live stream. I think uh I think you know for a lot of folks uh you know have used Replet as an app builder um and we’re adding additional functionality. So I think this is a was a really nice

1:48:59 introduction for folks that are new to Rep or even have been building on Replet for a while and just been using primarily to build apps in terms of what additional stuff you can do in Replet and also just like an overview of what it means to scale your efforts in this new AIdriven knowledge worker type of role. You know I keep going back to this analogy of the 100x engineer. I think it hit engineering early but I think it’s going to impact every other row out there around like moving up the value chain. So just you know see a task, automate

1:49:30 the task and then move up and find out what was the what’s the real value of the entire flow. Your systems approach thinking to this Rhar is really good. Like if you’re going to see the overall value of something it’s good to see where it fits in the overall flow of things and maybe like get at the tail end like why do people really need this or what’s the value of this and like try to get there. And then I also go back to this idea of being a full stack builder or knowledge worker. So whereas before you know you were maybe responsible for like getting features and collecting them getting feedback now you get to build now you get to go and find like

1:50:02 what is the right go to market approach and get actual users and then like do like things that maybe took a you know five person team uh to do now you get to all do that powered by tools like replet so >> I like full stack knowledge work worker >> full stack knowledge worker that’s generation >> you just you just dropped that in your uh and I think you glazed over it just like uh like kind of a kind of said it like quickly. Um but I think when you when you think of somebody as like a full stack engineer, they are somebody who can deal with the backend with the

1:50:33 front end. They can kind of look at security. They can look at the whole picture. They can look at optimizations and deployments and kind of like see the whole thing. And exactly what you said right there. If you’re a full stack knowledge worker, you have the ability to take an idea all the way from this little seedling into something that is useful in the real world, right? Whereas before a knowledge worker might have been somebody who suggested this concept and then you had to go find the engineer and then you had to go get the marketing resources and you had to go do whatever. No, no, no. The full stack knowledge

1:51:03 worker now is the person that has a little bit of the engineering knowledge, a little bit of the development slashdistribution knowledge. you have a little bit of the marketing insights, you understand the idea, but you can also walk that thing through into like reality. And I think that, you know, like that full stack knowledge worker is like that’s a nice little concept right there. >> Oh, I like it. Full stack knowledge worker. Let’s go with it. >> Send it, clip it, run it. [laughter] Okay. [gasps] >> Uh nice comment here from um

1:51:35 >> Oh, sorry. Let me get that off the >> Thank you so much. Appreciate it. No, thanks for hanging out with us this at here at the tail end of things. Um, >> appreciate all the love. >> Yep. Uh, let’s see. My auntie wants to know where to upload her API keys. Files. [laughter] Yes. Tell auntie to to not just, you know, don’t paste them in the chat. Just uh ask replet and replet will give you a nice little field to drop those in there.

1:52:05 >> Nice little prompt. You know, just drop it in here, auntie. Copy paste it here. Nowhere else. [laughter] >> All right, let’s see. Yeah. Okay, I think that’s uh Replet is the everything app. 100%. >> 100%. >> That’s nice. >> Feel the dance coming. >> Warm up. >> Another great value session. Yes, it’s getting really interesting. I haven’t done a lot of financial research with Rep. But yes, yeah, I think this is one of those things where you can kind of push yourself and expand a little bit on

1:52:35 what you think you [snorts] are able to do, right? If you have the idea for how to guide it, you know, you can kind of uh get some really interesting things pulled together. So, uh let’s see. All right, I think that’s um I think that’s it. Let’s let’s let’s do this thing. Oh. have a little dance off. Guys, you guys got to join us at home.

1:53:05 >> Her dancing and joining us at home. Let’s see those. Come on. >> We’ve been here 2 hours. Got to stretch it out, you know. >> All right, everyone. Drop your love in the comments. We’ll pull those on [music] screen. Thanks everyone for hanging with us again. We have office hours tomorrow in [music] Discord. We’re talking about how to price your repa projects. And then Friday, we’re back with another community showcase. [music] Um, stay tuned. Head over to community uh replet.com/ community for all the details about what’s happening in the space. And uh, [music] yeah, we’ll see you on the next stream. Bye everyone.

1:53:39 [music] [music] >> [music] >> They love you. [music] They love your dance, Manny. Non-stop. >> Two hours here, dude. I got to shake it out, man. Get loose. >> Oh, dang. Sorry. Crashed on the couch.

1:54:09 Hope you’re feeling better. Epic. >> All right. Thanks for tuning in, dude. Everyone, you guys [music] all rock. Thanks for hanging out. We appreciate everyone. We will catch you on the next stream. >> [music]