Claude & ChatGPT’s NEW Upgrades Change EVERYTHING! (Intelligent UI, Managed Agents & More)

Author: Rob The AI Guy · Watch on YouTube

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: Claude & ChatGPT’s NEW Upgrades Change EVERYTHING! (Intelligent UI, Managed Agents & More) · Channel: Rob The AI Guy · 12:24 · Watch on YouTube

Topics: AI agents, Claude, Prompt and context engineering, Software engineering with AI, AI models

Summary

The video reviews recent ChatGPT and Claude updates, including faster model options, more visual and interactive ChatGPT responses, and Claude’s ability to coordinate multiple agents. It also covers audio uploads, Claude’s model lineup, managed agents, live dashboards, animated explainers, and the new Reflect feature. The presenter highlights practical uses and cautions, including that managed agents can become expensive and that some features require specific settings or plans.

Key points

ChatGPT’s GPT 6.1 Soul Ultra-Fast is described as eight times faster than the regular version and is available through the API, Codex, and ChatGPT work.
ChatGPT’s model selector has been upgraded, and the video encourages users to switch to GPT 6.
ChatGPT can now respond to everyday questions with visual interfaces, such as diagrams, maps, learning graphics, and interactive tools.
The presenter says GPT-6 instant delivers responses 44% sooner than GPT-5.6 in instant search production traffic.
ChatGPT now supports audio uploads up to 512 megabits, but this feature is not available on the free plan.
Claude’s model selector includes Opus 5.5, Sonnet 5.5, and Haiku 5.5; the presenter recommends starting with Sonnet and moving to a more capable model if needed.
Claude managed agents can divide work among multiple agents in parallel and combine their results, but users should set a spending budget.
Claude can create live dashboards and animated explainers; dashboards can connect to services such as BigQuery, Databricks, Snowflake, and Salesforce.
Claude’s Reflect feature reviews recent usage and suggests prompts, automations, and ways to build skills.

Chapters

Start	Chapter	Summary
00:00	ChatGPT Ultra-Fast and model updates	The presenter introduces GPT 6.1 Soul Ultra-Fast, its availability and speed, then explains the updated ChatGPT model selector.
02:04	ChatGPT’s intelligent, visual UI	Examples show ChatGPT creating visual explanations, plans, and interactive experiences, alongside a faster response claim and expanded audio uploads.
05:40	Prompt tools and Claude’s model lineup	The presenter promotes a prompt optimizer and reusable shortcuts, then compares Claude’s models and recommends starting with Sonnet.
07:45	Claude managed agents	Claude can coordinate multiple agents to work in parallel, with a warning to set a budget because agent sessions can become expensive.
09:48	Dashboards, motion explainers, and Reflect	The video covers Claude’s live dashboards, animated explainers, required settings, and the Reflect feature for reviewing usage.
12:24	Closing	The presenter recommends another video for additional ChatGPT and Claude updates.

Quotes

ultra-fast right here is actually eight times faster than regular GPT 61 Soul. — 00:30

you could literally upload a photo of a bill and say create a bill splitter for a party based on what we ate. — 04:07

They then reconcile the findings and the results come back to the agent combined. — 08:17

you want to make sure that you set a budget to cap each session’s spend — 08:48

this is just scratching the surface of how this can actually help improve me using Claude. — 11:53

Concepts and entities

ChatGPT, GPT-6.1 Soul, GPT-6 Instant, My Prompt Buddy, Claude, Sonnet 5.5, Opus 5.5, Haiku 5.5, agentes gerenciados, Reflect

Full transcript (original language)

00:00 Chat GPT and Claude just released a bunch of brand-new updates that are going to change how you use these tools forever because now Chat GPT has an ultra-fast mode. In addition to that, they launched an intelligent UI and Claude just launched manage agents that completely changes what’s happening in the background when you’re using Claude. And by the end of this video, you’re going to know what all these changes are and how they’re actually going to affect how you use these tools. So, the first change that we need to talk about is Chat GPT has now rolled out ultra-fast for GPT 61 Soul. We could see right here

00:30 in terms of pricing and availability that GPT 61 Soul is now available to everybody, but ultra-fast right here is actually eight times faster than regular GPT 61 Soul. And GPT 61 Soul is pretty fast as it is. Now, this is now currently going to be available from inside of the API, from inside of Code X, and from inside of Chat GPT work. And if we come over here, we could see how this actually compares in terms of price and in terms of speed. So, we could see right here basically how this breaks

01:01 out. And if we actually come over here and pull up Chat GPT, I want to show you what this actually looks like from inside of here. So, let’s say that we come over into Chat GPT work. We could come over here into GPT 61 Soul. Now, if we come back over here, we could see how this actually adds up. So, we could see GPT 61 Soul standard. This is going to be just at medium. GPT 61 Soul fast is going to be, if we actually come over here and we change this, right? So, this is going to be us actually changing this and we could see we could go all the way up to ultra right here. And if we look

01:32 at this, this shows how much faster it is in terms of token generation with the model baseline and it shows that the price difference is. So, if speed and having really, really high intelligence really matters to you, this is going to be something you’re going to want to use. But for most of you, if that isn’t going to be something you’re focused on, you’re probably not going to want to use it, which leads me into update number two, which is something you were definitely going to want to use. So, now when you come over to ChatGPT chat right here and you come over and actually choose what model you want to use, we could see this is now been upgraded to

02:04 GPD6. In the past, this was GPD5.6. So, we could see GPD5.5 is going to be leaving in just 4 days. So, every single one of you that is using ChatGPT, make sure you’re using GPD6 and what you’re also going to notice now is that you now have an intelligent UI from inside of ChatGPT and this is pretty crazy here. For example, let’s say that we come over into ChatGPT and we ask it to break down the design of a seven-speed bicycle. This is what it’s now going to look like. We could see it’s now going to

02:35 give us an image and if we actually go through this, it literally breaks each of these down. If we want to see the wheels, highlights just those, tells us why it actually works. If we want to see the cockpit, highlights that, tells us what is actually inside of it. And in addition to that, this is actually way faster. So, now ChatGPT is going to make visual responses to everyday questions. For example, if you you tell it you’re planning a Sunday roast, on the left-hand side, this is what 5.6 instant would look like. Kind of ugly, a lot of text. It makes my ADHD go nuts because I don’t want to sit here and read any of

03:06 this. But now with GPD6 instant, this is what this is going to look like on the right-hand side right here, which is pretty cool. It actually builds things that we can use inside of here. And as we could see, it’s not just for something simple like that. You could get this to also plan a road trip where it literally goes through all of the photos, pulls up a map, everything that you should do. Absolutely incredible. In addition to that, this allows you to actually learn complex topics a lot more easily. Why? Because it actually builds graphics for all these. For example, let’s say that we came over here and we could say something like, I want to

03:37 learn more about compounding interest. Can you help me understand it better? Guess what? This is now going to be able to go through and the result that we get here, one, is not only way faster, but two is going to be much more visual. Literally created the chart inside of here and this literally goes through and shows me all of these different concepts and all of these different things that are going to be relevant to somebody that wants to actually learn about this. Again, incredible. And if we come down here, we

04:07 could see you can also now create interactive experiences just by asking. My personal favorite with this is probably this bill splitter right here. You could literally upload a photo of a bill and say create a bill splitter for a party based on what we ate. Can we make it look like a clipboard? And look at what this actually generates inside of ChatGPT. This is nuts. I mean, in the past this would have been an app that you would pay for. This is an app that some of you are probably still paying for, but now it’s literally just made directly inside of a ChatGPT chat and you could share it with other people. In

04:38 addition to that, as we could see right here, it delivers answers way quicker. This is what 5.6 instant looks like versus GPT-6 instant. It is way more in-depth. It is way more complex in a good way. Breaks things down visually and it just way faster, like drastically faster. In fact, if we look right here in terms of instant search production traffic, we could see that GPT-6 actually delivers things 44% sooner than GPT-5.6, which again is just incredible. And then that last update that I did

05:08 want to cover just very quickly for ChatGPT is that ChatGPT now takes audio uploads of up to 512 megabits, but it’s not available on the free plan. Essentially, what this means is you can now upload transcripts, audios, whatever, and you could pull stuff from it. You could have it create stuff from it. It’s pretty cool that it’s now able to do this and I don’t know why it took so long for this to happen. Now, before we get into the rest of those new updates, I did want to share with you something that ChatGPT still can’t do, which is that you still from inside of ChatGPT can’t save prompts that you use

05:40 over and over again and it doesn’t optimize your prompts for you. And one thing that you’re going to notice if If using GPT-6 Astra or ChatGPT’s latest models, is that the prompt that you give it drastically changes the output you get. And that is exactly why I use this My Prompt Buddy prompt optimizer right here. This allows you to insert all of your different prompts right here from a standard prompt, reasoning prompt, deep research, AI video, AI image. You could put in your current prompt, your ideal output. And as we can see from my history right here, this thing goes through and actually changes your prompt

06:10 so that you get a professional prompt right here that is much better. In addition to that, if we come over into shortcuts this is where you can actually save prompts that you reuse over and over again. So, you can simply just copy, paste it directly in here, and you don’t have to do anything other than that. And you don’t even have to leave ChatGPT in order to do that. It works with ChatGPT, Claude, and Gemini. So, it works with all the LLMs. And this is a no-brainer tool that I use every day. That’s exactly why I strongly suggest you go to the pinned comment below and get started with My Prompt Buddy today. I use this tool literally every single day. It’s compatible with Gemini,

06:41 ChatGPT, and Claude. And it’s going to make your life way easier. part about it is if you go to the pinned comment below, you can get started with a free trial today to test it out for yourself. Now, Claude update number one is now we are going to see that we have brand new models inside of here. If we now click into the model selector on Claude, we’re going to see that we have Opus 5.5, we have Sonnet 5.5, we have Haiku 5.5. I just had a member in my school community ask when they should use Haiku versus Fable 1 versus Opus versus Sonnet. And here’s how I would really break this

07:12 down. I would start everything with Sonnet 5.5. If it’s not able to handle it, I would go up to Opus 5.5. If it’s not able to handle it, I would go to Fable 5.1. But in terms of using Haiku, I personally never find myself actually using this because if we actually come over here and look at this, we could see that yes, it’s way cheaper. Yes, it’s fast. But if we actually look at this, it’s kind of terrible compared to everything else. I mean, it is slightly better than GPT-6 Luna, but I don’t find myself using that. If we compare it to Sonnet 5.5 for reference, we could see

07:45 like it is significantly worse. Like Agentic coding, it’s half of Sonnet 5.5. Even on knowledge work, 200 points less. Is it better than the old Haiku model? Yes. So, if you were using Haiku 4.5, I would begin to use Haiku 5.5, but quite frankly, this is probably a model that I just think you shouldn’t be using. Now, the next update I wanted to share here is going to be managed agents. So, if you have something that one agent won’t be able to do, now Claude will actually go through and create multiple different agents. So, check this out. You now have

08:17 all of these different agents working on all these different projects in parallel right here. They then reconcile the findings and the results come back to the agent combined. So, this will now literally go through if you have 300 contracts and be able to do it incredibly quickly. What this is going to do is one, make your Claude way faster, and two, just allow you to get way more done in a much quicker manner. One thing I do want to call out though, cuz as we could see that this is kind of calling it out at the end here, is you want to make sure that you set a budget

08:48 to cap each session’s spend because this can get incredibly expensive. Think about it. If you have 300 different things that you were trying to analyze and Claude says, “Hey, we need 100 agents to be able to do this.” That’s going to cost an insane amount of money when you have 100 different agents running in parallel, you might have just been better off having less agents do the task. Now, in terms of actually getting started with this, you’re going to want to make sure that you are coming over to Claude Code in order to do this. If you want to do an interactive

09:18 walk-through of this, you could literally come over here. Just take this right here and put this inside of the latest version of Claude Code and it will actually go through walking you through getting these agents set up. You are going to be able to set up the agent, the environment, you could specify the session and the events. In addition to that, if we come down here, this literally goes through all the prerequisites that you need and you you will be able to find this at platform.trawld.com/docs/en/manage-agents or if you just do a Google search for

09:48 manage agents, you’ll see it. Now, that last Trawld update that I did want to cover is you can now build live dashboards and animated explainers with Trawld. So, this is going to now be rolled out. We could see that docs, slides, and designs are now out of beta on every Trawld plan including free. So, starting today, you can now turn your data into a dashboard that stays current and your ideas into animated explainers for presentation, which is pretty cool. In addition to that, we could see that Trawld dashboards does connect to

10:19 platforms like BigQuery, Databricks, or Snowflake, or CRM tools like Salesforce. Using it is incredibly simple. You’re literally just going to come over here and say make a dashboard artifact for whatever type of data you want. You either then have to give it access to data or give it access to the connectors so that it will be able to access the data that you actually want it to show. And for Trawld motion, what you’re going to want to do is come over here, type in {slash} motion, and then describe exactly what you want to actually build. And we could see right here, Trawld motion is built for explaining your own

10:51 work. So, Trawld writes the code that actually animates your text, your charts, your shapes, your images so that you could change any word, numbering, or timing. It doesn’t use a video generation model, so there’s no generated footage and no AI generated people. And when you want to take it further, you can actually open it inside of something like Adobe, Descript, HeyGen, Hixfield, or any video editor inside of it. And then you can actually change things or customize them inside of there. And in order to make sure that these things are actually turned on, you want to make sure that when you come

11:21 into your settings on Trawld, you come over into capabilities, tool access, connector search, switch models when a message is flagged, artifacts, AI-powered artifacts, and inline visualization. You want to make sure that all of those are going to be turned on. Because if they’re not turned on, guess what? Claude isn’t going to be able to do this. In addition to that, we also see the reflect feature now inside of Claude. So, what this actually does is goes through our last month with Claude and suggests different ways for us to use it, different automations that we should set up, and more. This is a

11:53 feature we’ve been waiting on for a long time, and it looks like it’s silently has now been rolled out. And we can see that this little image now goes through, tells me everything that I did here, shows me different ways for me to actually expand my skills, and different things that I should do, gives me prompts, and this is pretty cool. The only thing I wish is that this gave us a lot more information and was way more in-depth. Because this is cool, but I think this is just scratching the surface of how this can actually help improve me using Claude. Now, if you enjoyed this video and you want to learn more about the latest ChatGPT and Claude

12:24 updates that I didn’t cover in this video, I would strongly suggest you check out this video right here that goes through all of them. I’ll see you over there.