Episode Details

Hey folks, Alex here, writing from... a undisclosed tropical paradise location 🏝️ I'm on vacation, but the AI news doesn't stop of course, and neither does ThursdAI. So huge shoutout to Wolfram Ravenwlf for running the show this week, Nisten, LDJ and Yam who joined.

So... no long blogpost with analysis this week, but I'll def. recommend tuning in to the show that the folks ran, they had a few guests on, and even got some breaking news (new Flux Kontext that's open source)

Of course many of you are readers and are here for the links, so I'm including the raw TL;DR + speaker notes as prepared by the folks for the show!

P.S - our (rescheduled) hackathon is coming up in San Francisco, on July 12-13 called WeaveHacks, if you're interested at a chance to win a RoboDog, welcome to join us and give it a try. Register HERE

Ok, that's it for this week, please enjoy the show and see you next week!

ThursdAI - June 26th, 2025 - TL;DR

* Hosts and Guests

* WolframRvnwlf - Host (@WolframRvnwlf)

* Co-Hosts - @yampeleg, @nisten, @ldjconfirmed

* Guest - Jason Kneen (@jasonkneen) - Discussing MCPs, coding tools, and agents

* Guest - Hrishioa (@hrishioa) - Discussing agentic coding and spec-driven development

* Open Source LLMs

* Mistral Small 3.2 released with improved instruction following, reduced repetition & better function calling (X)

* Unsloth AI releases dynamic GGUFs with fixed chat templates (X)

* Kimi-VL-A3B-Thinking-2506 multimodal model updated for better video reasoning and higher resolution (Blog)

* Chinese Academy of Science releases Stream-Omni, a new Any-to-Any model for unified multimodal input (HF, Paper)

* Prime Intellect launches SYNTHETIC-2, an open reasoning dataset and synthetic data generation platform (X)

* Big CO LLMs + APIs

* Google

* Gemini CLI, a new open-source AI agent, brings Gemini 2.5 Pro to your terminal (Blog, GitHub)

* Google reduces free tier API limits for previous generation Gemini Flash models (X)

* Search Live with voice conversation is now rolling out in AI Mode in the US (Blog, X)

* Gemini API is now faster for video and PDF processing with improved caching (Docs)

* Anthropic

* Claude introduces an "artifacts" space for building, hosting, and sharing AI-powered apps (

Listen Now