Inline Charts, New Models & Jev | This Week in Replit (Oct 2, 2026)

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: Inline Charts, New Models & Jev | This Week in Replit (Oct 2, 2026) · Channel: Replit · 05:39 · Watch on YouTube

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

Summary

This week in Replit highlights interactive charts in chat, enabled by the acquisition of Ada, and encourages users to bring their own data or connect data sources. The discussion also covers new models available in Replit, including GPT-6.1 Soul, Claude 3 5.5, and Jev, and explains how users can select models or let Replit choose in auto mode. Jev is presented as a useful option for classification tasks and is available through Replit’s AI integrations. The episode closes by previewing upcoming conversations with Replit team members about authentic selling and a journey from teaching to building and teaching AI.

Key points

Users can create interactive charts in chat using public data, their own data, or connected data sources.
The charting functionality is associated with Replit’s acquisition of Ada; its founders, Omar and Amin, are scheduled to appear on the show.
GPT-6.1 Soul is available in the model picker and is described as suited to tougher tasks in max mode.
Claude 3 5.5 is described as a cost-efficient model for everyday tasks.
Users can select a model for a build or use auto mode to let Replit choose.
Replit’s model orchestration lets users access and switch among models with different strengths.
Jev, released by AI lab Type Safe, is described as a good option for classification tasks such as sorting emails or support tickets.
Jev is available through Replit AI integrations, which pull it from OpenRouter and bill through Replit without requiring a separate API key in the environment.
The show previews a segment with sales team member Andrew about authentic selling, followed by Noni discussing her path from teaching to building and teaching AI.

Chapters

Start	Chapter	Summary
00:00	Interactive charts in chat	The hosts introduce chat-based interactive charts, discuss bringing in data, and highlight the functionality’s connection to the acquisition of Ada.
01:33	New models and Replit’s orchestration	The hosts discuss GPT-6.1 Soul and Claude 3 5.5, along with choosing models or letting Replit handle selection in auto mode.
03:37	Testing models and introducing Jev	After describing how to try models for different tasks, the hosts introduce Jev as a classification-focused model available in Replit.
05:09	Jev and upcoming guests	The hosts briefly discuss Jev’s architecture and preview conversations about authentic selling and a teacher’s journey into building and teaching AI.

Quotes

Now, you can build beautiful and interactive charts in chat. — 00:00

It’s like now everybody gets to be a data person. — 01:02

you get to use all the models here, right? — 02:04

Just ask Replit. That’s it. — 03:37

Jev is the first model released by a new AI lab company called Type Safe — 04:08

Concepts and entities

Replit, Ada, GPT-6.1 Soul, Claude 3 5.5, Opus 5.5, Jev, TypeSafe, OpenRouter, Harness, Omar, Amin, Rima

Full transcript (original language)

00:00 Let’s talk a little bit about what’s happening this week in Replit. Excited about this update. Now, you can build beautiful and interactive charts in chat. You can grab publicly available data. You can bring your own, or you can plug into your favorite data sources and ask Replit to build out these charts for you. This functionality is coming to us via the acquisition of Ada, and the founding team, Omar and Amin, are going to be on the show in a couple of weeks

00:30 to tell us all about it. But in the meantime, go into Replit, bring your data, and bring these beautiful visuals to life. We did a deep dive this past Tuesday. >> Yeah. >> Rima worked through a couple of examples, and we clipped some of that stuff, and people really like it. So, really cool things that you can do now here in chat. >> Yeah, and chat’s just kind of really pushing the limits what you can do inside of there. And so, go check out the stream we did Tuesday where we walked you through routines, these charting, all the different stuff you can do inside of chat, and just thinking

01:02 about knowledge work in general. But this is a big boost right here having the ability to like pull these reports, work with data right in the chat. So, yeah, super cool to see this coming through. >> I had a chance to talk to Omar and Amin last week, and I think they framed it perfectly. It’s like now everybody gets to be a data person. As a as a former data professional, I can tell you the stuff that you can do now in chat is like stellar. It’s so good. So, definitely take the time to go play around, just get familiar with that functionality, and bring your data to life.

01:33 Another week and more models dropping, but good news, they are available in Replit. Now, we have OpenAI’s GPT-6.1 Soul that’s available in our model picker. The buzz around this model is that it’s really good for your tougher tasks. So, this is going to be available in our max mode. And then also dropping this past week was Anthropic’s Claude 3 5.5. Really good for everyday tasks. It seems to be a really good cost-efficient

02:04 model. So, you can go and select those models for your builds, or you can always stay in auto mode and let Replit choose the best model for the job. >> Somebody asked the other day, we were in a in a call and they asked a little bit about like, what makes this different than if I am using you know, a Claude code or a Codex or or one of those things? And the answer was like, you know, you get to use all the models here, right? Like one of the cool things about Replit and where we’re making our bones is in this agent orchestration layer, this harness, you know, this agentic harness you’ve heard people talking about. And so, having access to, you know, all of the cutting-edge models

02:36 as they come out, being able to choose and select between them, if you know a little bit about that, and being able to like tune that how you like it, or being able to like Manny said, put it into auto mode and just kind of let that thing roll on its own, I think it’s just a such a powerful advantage, and we’re going to see that become a bigger and bigger advantage as these different models mature. Not to mention that sometimes the different models do different things well. So, like maybe some of them do 3D modeling, whereas some of them do design, whereas some of them do you know, like security or you know, the code generation itself. And so, like being able to pass it off

03:06 and kind of like have the best of all the different models is really what makes Replit shine. And I think again, having these cutting-edge models always just available uh just a a testament of how powerful and diverse the the tool is. >> Mhm. Yeah, I think more and more people are going to be working in auto mode, and um we actually dropped a blog post on that that I think we’ll get to towards the later part of the show. But, like for example, um yesterday I wanted to put together in an explainer video on Jev, and I I’ve been hearing that Opus

03:37 5.5 is really good at video animation. So, I just selected Opus 5.5 and got like a really good video out of that. So, I think it’s a great um way to test out different models to different functionalities based on what you’re seeing out there as they’re released. So, best of both worlds, you know, if you want to try something out, you want to try close source model and open source model, you can do that. Or if you just want to stay in auto mode and just stay in flow, let Replit take care of the heavy lifting, you can do that, too. >> Just ask Replit. That’s it. >> [laughter] >> Go back to Tuesday show, folks. That’s

04:08 very bingeable, that show. Speaking of new models, we also have Jev now available in Replit as well. Jev is the first model released by a new AI lab company called Type Safe, and a really good model for classification problems. So, if you’ve got a bunch of emails that you want to classify as spam or not spam or categorize, or you’ve got a bunch of support tickets that you want to quickly go and filter and categorize, Jev seems

04:39 to be a really good model for that. It’s now available in Replit through our AI integrations functionality, which means that you can just call it from Replit, and Replit will pull that in from OpenRouter and pass the billing directly to you in Replit. So, you don’t have to go and get an API key and bring that into your environment. You can continue to build and still leverage this model and any other model that is on OpenRouter router, as well as models from Anthropic, OpenAI, or Google Gemini, bringing those into your

05:09 applications. There’s been a lot of buzz around Jev, and I’ve been digging in on Jev. Really cool architecture, really nice new option for certain types of use cases. So, we’re going to chat a little bit more about Jev a little bit later in the show. >> All right. >> Yeah, nerd out a little bit here. >> [laughter] >> Nerd mode activate. All right. >> But before we go nerd mode, we got to have a little fun, and for that, we’ve got Andrew coming up in just a bit. He

05:39 is on the sales team here at Replit, and he’s going to talk to us about how to become an authentic seller. I saw him coin that phrase on LinkedIn. I’m like, yes, that’s it. That’s where people are going. Like they We talked about the full stack knowledge worker and now we’ve got the energetic seller here. So, Andrew’s going to tell us all about that and then after that we’re going to have Noni, one of the staples in our community to talk to us about her journey of going from a teacher to becoming a builder to becoming someone that’s actually going out and teaching people about AI.