ThursdAI - The top AI news from the past week · From Weights & Biases, Join AI Evangelist Alex Volkov and a panel of experts to cover everything important that happened in the world of AI from the past week

📅 ThursdAI - Feb 19 - Gemini 3.1 Pro Drops LIVE, Sonnet 4.6 Closes Gap, OpenClaw Goes to OpenAI

·1 hr 32 min·5 clips

Gemini 3.1 Pro is live in Google AI Studio, and Noam Shazir called it breaking news.

1. ThursdAI’s February 19 episode centers on Gemini 3.1 Pro, OpenClaw, Sonnet 4.6, and the week’s AI release cycle. 2. Alex Volkov hosts the panel, and Wolfram Ravenwolf, LDJ, Ryan Carson, and Jan Pelleg contribute benchmark takes, agentic-coding opinions, and product-use experience. 3. The episode asks what matters most in a week where Google, OpenAI, Anthropic, Alibaba, ByteDance, and Google DeepMind all ship new models or tools. 4. Alex Volkov opens with Gemini 3.1 Pro in Google AI Studio and says Noam Shazir posted that it is the breaking news. 5. The panel lists benchmark numbers for Gemini 3.1 Pro, including 44% on Humanity’s Last Exam, 77% on RKGI, and 68 on Terminal Bench. 6. LDJ highlights Zephyr’s ZUNA, a sub-billion-parameter model framed as “thought to text” for EEG signals from the brain. 7. The TLDR rundown names Qwen 3.5, a 397 billion parameter open-weights multimodal MoE, and Cohere Labs’ tiny AYA 3.3 billion multilingual family. 8. The same rundown also mentions ByteDance Seed 2 as a frontier multimodal LLM and Google DeepMind’s Lyria 3 music-generation model in Gemini. 9. Jan Pelleg and Ryan Carson focus on OpenClaw after Peter Steinberger joins OpenAI in what the panel calls a founder acquisition. 10. The discussion adds that OpenAI lets Pro users use OpenClaw, while Anthropic’s terms and later clarifications become part of the debate. 11. Wolfram describes a Cloud Code meetup where many attendees still were not running agents for themselves. 12. Jan Pelleg says people in his local Llama subreddit had become unusually negative about OpenClaw and were questioning whether it was real or useful. 13. Ryan Carson says his group of older coffee companions felt fear and dread, then he told them to sign up for paid accounts on Gemini, Cod, or ChatGPT. 14. Jan Pelleg says OpenAI’s documentation on building a “CodeFactory” with Codex pushed him to set aside two or three days and rework his repo. 15. Ryan Carson argues that harness engineering matters because the lab that builds the model is also best positioned to build the harness around it. 16. The panel repeatedly compares Codex, Cloud Code, Opus 4.6, and Sonnet 4.6 as products that combine models, harnesses, and prompt workflows. 17. The tone is fast, conversational, and reactive, with overlapping speakers, side comments, and benchmark screens shared live. 18. The format mixes a weekly news rundown with live commentary, product testing, and argument about how agents are actually used. 19. People tracking AI releases, coding agents, and model benchmarks will get the most from it. 20. Listeners wanting a calm overview without live benchmark debates may skip it.

As heard by us

A live AI news sprint where Gemini 3.1 Pro and Sonnet 4.6 meet the software hiring question.

Gemini 3.1 Pro lands just as ThursdAI goes live, giving Alex Volkov and the panel a chance to react while the release is still unfolding. Volkov, the host and an AI evangelist with Weights and Biases, is joined by Wolfram, Raven Wolf, and LDJ before the conversation moves from…

Read the full review in PlayNext →

Why you'd press play

Want Gemini 3.1 Pro, Sonnet 4.6, and agentic software hiring anxiety in one caffeinated briefing?

Read the full recommendation in PlayNext →

Listen to the show on

Apple Podcasts Spotify

Play clips from this episode More from ThursdAI - The top AI news from the past week