POD UP! A fresh ThursdAI for Oct-16 with tons of content! We had 4 great interviews, PM for VEO3.1 (Jessica), @swyx with SWE Grep breaking news, @corbtt talked to us about RL and @sqs about ad powered free coding agents! 👉 thursdai.news/oct-16 👈

POD UP: Covering my 3rd @OpenAIDevs Day in a row, and this one includes a few questions from me to @sama and @gdb from an exclusive fireside chat + a full breakdown of what they ⛴️, interview with @pvncher, Samsungs 7M TRM & building agents with AgentKit, Sora vs Grok Imagine & much much more! 👇

I don't remember when I had so much fun on the stream! This week on @thursdai_pod we started an invite chain for SORA2 + shared tips and tricks, covered the product and the AI aspects! We also covered Claude 4.5, @Zai_org GLM, the DSA innovation from @deepseek_ai and much more!

Sora2 from @openai released just 2 days ago, and people are creating and remixing incredible things. The top video with 2.2K likes is a selfie video of Jesus during the last supper, but it only starts there. Sora allows folks to remix any video by adding their own prompt.

Very welcome addition, the newsletter website has support for AB testing titles! Very interesting that the title that got the most opens for my last @thursdai_pod was the one with the 📆 emoji but without the date + a little bit of a clickbaity title 🤔 Will keep testing 🤔

Holy Sh*T - I swear to god the weeks are getting denser! Somehow we still manage to squeeze the news about NVIDIA $100B, OpenAI Pulse, @Alibaba_Qwen total OSS domination, new @Kling_ai @SunoMusic V5, new evals and benches and a LOT more in under 1.5 hours of @thursdai_pod 👇

These weeks are getting denser and denser! We covered @openai Codex, ICPC updates and Usage paper, @Meta display glasses, @reve incredible UI and Model, @LumaLabsAI HDR Ray3, chatted with @jerber888 about @arcprize SOTA with Grok, @theworldlabs Marble and SO MUCH MORE 👇

Wohoo! Made it! Edited this @thursdai_pod episode across 2 states, in ubers, airplane and with 3% battery - it's ready on time on Thursday! We covered a LOT of ground, so this one is worth watching as well as reading. Huge thanks to everyone who contributed, tuned in, shared and open sourced their

My team ran benchmarks using realistic LLM training workloads (llama 70B). CoreWeave infra is the fastest and most stable in the industry. Large distributed jobs survive longer w/o interruptions and recover from faults orders of magnitude faster, reducing wall time, GPU hours, and $$ to train. 1/

Wohoo, I'm back in the studio today, and what a great @thursdai_pod today! We had a lot of fun talking about Codex, Hermes, @OpenPipeAI acq, Antropic nerfing Claude, Gemma Embeddings and a lot more! Feels so good to be back! Links in the next post (Until Nikita fixes the links thing)

Quick one this week, we had a quiet-ish week until DeepSeek dropped V3.1 (which we covered) + a lot of new image editing models, both open source ones and .. 🍌 Enjoy this episode if you didn't tune in live 👏 👇

Phew, another @thursdai_pod dropped. This week was a bit more "chill" and yet we still had a LOT to talk about, 5 open source models (with breaking news), 2 open source world models + everyone spent a week with GPT-5 and we had thoughts! 👇

WHAT A SHOW! Folks, just sent the last @thursdai_pod and it's a "must not miss" episode! We had over 20K folks tune in to our livestreams today & yesterday and covered @xai Grok-4 (w/ controversies), @huggingface foray into robotics, had breaking news from @LiquidAI_ @RekaAILabs check it out 👇

🎙️ New @thursdAI POD - June 19 - we talked with @pankaj about @yupp_ai , reviewed 3 new video models from @Hailuo_AI ByteDance and @midjourney and covered a lot more AI What a great week for AI and a great conversation with @WolframRvnwlf @nisten @Yampeleg & @ldjconfirmed 👇

Had an absolute blast on @thursdai_pod today! Energy was flowing, we discussed just an incredible amount of news, while also covering 2 guests @pvncher (RepoPrompt) and my friend @Stefania_druga who helped cover @aiDotEngineer summit, and her incredible science copilots! 👇

This week on the pod! Live from VEGAS, we chatted with @LinusEkenstam and @charliebholtz about "a week after Opus and VEO3", Analyzed breaking news from Deepseek with @Yampeleg and @nisten and geeked out on AI Avatars with @WolframRvnwlf

Very much looking forward to this keynote presentation (alongside @LinusEkenstam @skirano @jerrod_lew @nkwrnr @dsiroker @bindureddy and others!) While preparing for this keynote, I supercharged my processes with a few new tools and AI flows.

x is dead? what's going on?

This week on ThursdAI - Gemini 2.5 IO edition, why we didn't love Mistral's new release, @WolframRvnwlf Qwen evals (it's good model ser) and a reality breaking AI Avatar tech from @HeyGen_Official that you can try today! (I gotchu all 1 month free code!)

Holy Shit! The AI Avatar space is heating up! @HeyGen_Official just added Avatar IV which only takes a pic and it looks ... REALLY good. Almost too good TBH. Face, emotions, hand movements, perfect lipsync, all from 1 photo! A new SOTA for sure! (A few comparisons in thread)

BREAKING: Microsoft as always silently drops Phi models and now they've added reasoning to them. Two models: Phi-4-reasoning 14B and Phi-4 Reasoning Plus (trained to output longer thoughts) Achieving a great 78 at AIME with 14B params is not bad! x.com/suriyagnskr...

Phew what an intense AI week folks! It's been getting crazier! Here's mostly human written ~3700 words and links from everything that happened this week + an edited version of @thursdai_pod just landed wherever you get yours! Links as always in... first comment

🚨 BREAKING: Google's response to o4-mini and 4.1-mini is here in one model! Gemini 2.5 Flash is announced in preview, and this one model can go from very cheap and fast, to a more expensive with reasoning, automatically and manually! Here's everything you need to know 👇

So... new model release day! Congrats to OpenAI for these awesome new models! I did a @weave_wb evaluation (on @AIExplainedYT 20 logic questions) of the new Quas... sorry 4.1 series of models, and compared with 4o and 4.5 and the results are super interesting 👀👇

Honestly didn’t think we’d make it to 100 episodes of @thursdai_pod — but here we are. From context windows jumping from 4K → 10M, to protocols, to chaotic Breaking News and model drops… this week had it all. Perfect way to celebrate. 🧵

with the new chatGPT enhanced memory that remembers all your chats, ask it this: > What's a thing that I've asked you that I likely forgot by now that you think is important and surprising? It will give you a few things that you may remember, but I definitely forgot this one

Honestly didn’t think we’d make it to 100 episodes of @thursdai_pod — but here we are. From context windows jumping from 4K → 10M, to protocols, to chaotic Breaking News and model drops… this week had it all. Perfect way to celebrate. 🧵

Ok new chatGPT memory prompt share > What is the most boring thing I've asked you I got totally cooked 😅 bruh

NEW: Proud to announce that @weights_biases is among the partners contributing to the new A2A (Agent 2 Agent) protocol that @Google just launched! A2A is a new protocol that allows for agents to discover other agents with specific capabilities via a registry! More Details 👇

Just published the #99's episode of @thursdai_pod 🤯 We chatted about multimodal SOTA embeddings with @zach_nussbaum , about training a SOTA model for Github Issues with @xingyaow_ from All Hands and covered this weeks tons of AI news, from OpenAI to Gemini 2.5 Check it out👇

🤯 I was asked internally to summarize @thursdai_pod in 1 slide and just noticed, that just after 2 years of the live shows, 1.5 years of podcast releases and almost 100 episodes (this week is #99!) we have just crossed... 100,000 downloads on Sub**ack 🔥

Honestly want to just pause for a sec and reflect A week like this week, where Gemini 2.5 broke all charts (closely followed by GPT 4o) & then 4o native image gen completely overtaken the internet is why we do @thursdai_pod in the first place! tune in with the links below

Gemini 2.5 pro "gets" me 😍

Focus mode in @raycastapp is better than Aderall.

Honestly want to just pause for a sec and reflect A week like this week, where Gemini 2.5 broke all charts (closely followed by GPT 4o) & then 4o native image gen completely overtaken the internet is why we do @thursdai_pod in the first place! tune in with the links below

Double return to interrupt is such an underrated UX pattern in AI chatbots, that after using it in @windsurf_ai , I want it everywhere! ChatGPT, every chatbot TBH should have this.

Had such an amazing time today on the show! Tons of guests, at least 3 breaking news segments including a live watch of the OpenAI new voice modes announcements. Tons of open source (@MistralAI, @TencentGlobal, @NVIDIAAI, Canopy) 👉 thursdai.news/mar-20 👈

Here's a snippet from my new live reaction video for the new OpenAI TTS model gpt-4o-mini-tts where I just told it how to pronounce ThursdAI (it's Thursd-ay-i folks, not Thursday AI 😅) and it just gets it! The whole release is insane (full YT link in comments)

Finally got around to record and edit the talk I gave at @aiDotEngineer summit NY a month ago. I wanted to explore whether reasoning models make better LLM judges. I had a lot of fun researching and giving this talk. Many asked for a recording / slides - YT link in comment

Alibaba's R1-Omni The first (?) application of Reinforcement Learning with Verifiable Reward (RLVR) to an Omni-multimodal large language model in the context of emotion recognition, a task where both visual and audio modalities play crucial roles.

Thanks to @ReidMayo finally got a few pics of the crowd from my Expo talk at @aiDotEngineer last week. There was a LOT of people, a lot of noise as well. I'll 100% record and post this talk soon (as soon as I get better, a sizable crowd is not the only thing I picked up there 🤧)

What a BANGER week that doesn't stop! From o3-mini to DeepResearch to OmniHuman to Gemini 2.0 to Copilot Agents and MUCH much more! Here's a supercut (made by AI) of the episode today, including quotes from @DeryaTR_ from our show today! The full links in comments as always x.com/thursdai_po...

This is crazy

ByeTok

Tons of new open source this week From new #1 TTS that sounds incredible (and brought back the sky voice from OpenAI!) to new MoonDream that can analyze where people look and new LipSync SOTA + Nvidia CES updates, OpenAI ASI ambitions & more AI news! Don’t miss this one! 🔗👇

Interested in AI agents, try this excellent CrewAI flavoured overview, with probing questions by @altryne.bsky.social incl shortcomings..lack of protocols for integration and scaling ThursdAI - Jan 2nd 2025 - Interview w/ Joāo Moura - is 25' the "year of AI agents"? www.youtube.com/watch?v=__22...

Notes for tomorrow ready! Banger week OpenAI unleashing updated o1 in API and it breaks all evals, Google is one-upping SORA with VEO2, you can now call chatGPT, a bunch of OSS releases, Genesys does gen physics⁉ & just tons of amazing stuff to chat about 👇

I know this word is overused but THIS WEEK is indeed INSANE. Gemini 2.0 took the 👑 but tied for 1st place with @OpenAI insane stream of releases from SORA to AVM with 👀 but not to forget we got 🦙3.3 and @xai image model! We somehow managed to cover all on @thursdai_pod 👇

🔥 LLama 3.3, SORA, new Deepseek, ChatGPT in IOS, , NeurIPS drama, and now Gemini-2 Flash!!! This may one of the biggest AI weeks of the year! thursdai.news/next Join us tomorrow for a ThursdAI recap!

Elon, add links to a comment, what's a big deal. X algorithm: LOL First post in a thread ~3.8K views Second post with links... drum roll... 78 whole views. yeah, it's not a typo, 78 views (not even people) so like.. 2%? Third post in the same thread... 1K views. My links were literally hidden 😂