Alex Volkov (Thursd/AI)

When we break through the coding singularity, it won't be immediately apparent but there will be signs! Autonomous agents running for 14h straight with tasks... 📅 Feb 26 Approaching singularity: Devin 2, METR 14h, Qwen 3.5, WarClaude, Distill attacks & more

Or on YT:
www.youtube.com/watch?v=4kW...

📅 ThursdAI - Feb 26 - Approaching singularity From Weights & Biases, this week is the closest I've felt to the AI singularity starting, bonkers 1 man AI startups crossing $700K ARR live on show, DoD vs Anthropic, Anthropic vs Chinese models & mor

You can also just listen to the show right here:

thursdai.news/feb-26

Philip Kiely - guest on ThursdAI podcast Philip Kiely (Baseten) on ThursdAI with Alex Volkov. Listen to episodes, find links, and explore the guest network.

And finally, @philipkiely was there with his first book! Inference is everything as Phillip said! Inference Engineering is available as a free PDF and as a gorgeous book (that I just also got in the mail!)

Don't miss these interviews 👀

thursdai.news/guests/phil...

Ben Broca - guest on ThursdAI podcast Ben Broca (Polsia) on ThursdAI with Alex Volkov. Listen to episodes, find links, and explore the guest network.

@bencera_ straight up gave us singularity vibes — dude crossed $700K ARR LIVE on the show

thursdai.news/guests/benc...

x.com/altryne/sta...

Nader Dabit - guest on ThursdAI podcast Nader Dabit (Cognition) on ThursdAI with Alex Volkov. Listen to episodes, find links, and explore the guest network.

Chatting with @dabit3 was amazing — been following his career forever

@nisten even said watching one of Nader's vids changed his whole career path!

Nader just joined @cognition and walked us through why

thursdai.news/guests/dabit3

📅 ThursdAI - Feb 26 - Approaching singularity — ThursdAI ThursdAI Feb 26: Anthropic vs Pentagon, METR 14.5h autonomy, Devin 2.2, Qwen 3.5, and Polsia's $700k ARR surge.

You can find the edited version of our live show, show notes and links on our brand new (totally not vibecoded) website here:

thursdai.news/ep/feb-26-2026

3 years doing this weekly and I've never felt closer to the singularity than right now

Everyone's shipping async AI agents. Everything's converging.

This week we covered Anthropic vs DoD, had 3 incredible interviews with @dabit3 @philipkiely and @bencera_ and way more 👇

This week on ThursdAI: open source models are closing the gap on frontier labs FAST — and we had the researchers behind them live on the show.We interviewed ... 📆 ThursdAI - GLM 5, MiniMax 2.5, Seedance 2, Gemini 3 Deep Think, Codex Spark 100tps & more AI news

Finally we live reacted to the drop of Cerebras powered Codex and Gemini beating Arc AGI and debated the AI psycosis which makes it hard for us to sleep!

Check out the full episode youtu.be/wQb4JK5xKMw

Then a chat with Olive revealed how the heck do they get close to Opus on Swe Bench verified with only 10B active parameters!

This was a packed show, Open Source LLMs are catching up, @louszbd from @Zai_org told us the new GLM 5 is for agentic architecture and is bigger, better, faster stronger.

📆 ThursdAI - GLM 5, MiniMax 2.5, Seedance 2, Gemini 3 Deep Think, Codex Spark 1000tps & something big is coming From Weights & Biases - 2 interviews with Lou (Z.AI) and Olive (MiniMax) about their SOTA OSS LLms, Seedance 2 shattering reality, Codex / Claude get faster and Gemini nearly beating Arc AGI 2

First of all, find the whole show on YT (link in bio) and here:

thursdai.news/feb-12

I knew we named it @thursdai_pod for a reason! AI labs LOOOVE dropping stuff on a Thursday!

Just today we had:

MiniMax 2.5 (+ @olive_jy_song hopped on)
OAI 5.2 Spark - @cerebras powered
and Gemini Deep Think beating Arc AGI 2 SOTA by 16 points

+ covered GLM5 and Seedance 👇

MiniMax went from relative unknown to frontier-competitive in ONE generation jump (M2.1 → M2.5).

Olive Song is joining @ThursdAI_pod LIVE today to discuss M2.5, the architecture, and what's next.


Don't miss it 🔥

Follow @altryne and @ThursdAI_pod

But THE chart that matters — cost vs quality:

M2.5 hits ~57% win rate at $0.15/task
Opus 4.6 hits ~73% at $2.40/task
GPT-5.2 hits ~55% at $1.75/task

M2.5 is 10-15x CHEAPER than frontier models for near-equivalent performance 💰

AGENTIC + TOOL USE:

• BrowseComp: 76.3 (beats GPT-5.2's 65.8, beats Gemini 3 Pro's 59.2)
• BFCL multi-turn: 76.8
• Wide Search: 70.3
• RISE: 50.2 (matches GPT-5.2)

Serious agentic capability here.

CODING:

• SWE-Bench Verified: 80.2 (vs Opus 4.6: 80.8, GPT-5.2: 80)
• SWE-Bench Pro: 55.4 (vs Opus 4.6: 55.4 — TIED)
• Terminal Bench 2: 51.7

This is a Chinese lab matching Anthropic and OpenAI on coding. Let that sink in.

BREAKING: MiniMax just dropped official M2.5 benchmarks and they're going HEAD TO HEAD with Opus 4.6, GPT-5.2, and Gemini 3 Pro 🤯

And Olive Song from @MiniMax_AI is joining ThursdAI LIVE in ~30 min to break it all down

@ThursdAI_pod

Here are the numbers 👇

WHAT a week for AI agents! Moonshot AI dropped a triple release — Kimi K2.5 with 1 TRILLION parameters achieving SOTA on agentic benchmarks, Kimi Code as an ... 📆 Jan 29 - Kimi 2.5, RIP Clawdbot, Anthropic Apps, Karpathy Says 80% AI-Coded & more AI news

You can join our live show right here on my X or on youtube:

www.youtube.com/watch?v=Nrm...

We don't do twitter spaces anymore because they suck but it was fun for a while!

Another MASSIVE week in AI 🔥

Moonshot dropped Kimi K2.5 1T — SOTA on agentic benchmarks, Anthropic shipped MCP Apps , Karpathy says 80% of his code is now AI-written, RIP Clawdbot and much much more!

All this and more on @thursdai_pod tomorrow at 8:30am PT!

Welcome to this week's ThursdAI for January 15th, hosted by Alex Volkov from Weights & Biases. This episode dives deep into the world of AI with a focus on a... 📆 ThursdAI - Jan 15 - Agent Skills Deep Dive, Claude Cowork for the Masses, GPT 5.2 Codex in the API

If you're building with agents in 2026 and haven't looked at skills yet, watch this one

Progressive disclosure >> dumping everything into context and praying

Link in first tweet ✨ and on YT here:
youtu.be/qszIEZzNRsI

The Eleanor convo was so good! A very clear communicator!

Skills are just markdown files but they solve the "how do i teach my agent MY workflows" problem so elegantly

I got so excited during pre recording i vibecoded skills into Chorus for ALL models not just Claude 😅

📆 ThursdAI - Jan 15 - Agent Skills Deep Dive, GPT 5.2 Codex Builds a Browser, Claude Cowork for the Masses, and the Era of Personalized AI! From Weights & Biases - come learn what agent skills are all about, Claude Cowork opens the door for non coders to do agentic stuff, GPT 5.2 Codex in API and Gemini get personalized! Big week!

We went DEEP on why skills are the missing piece for agent customization without destroying your context window

Also Cursor let GPT-5.2 Codex run for a week and it built a browser from scratch. 3M lines of Rust code 👀
thursdai.news/jan-15

🎙️ POD UP! holy shit what a week

Agent Skills finally clicked for me, Codex built a BROWSER in a week, Cowork is letting non-coders do agentic stuff, Gemini got creepily personal 🤯

We got @intellectronica on the show to break down skills and it was 🔥

📆 ThursdAI - Jan 8 - NVIDIA's Vera Rubin 5x Leap, XAI's $20B Raise, Ralph Wiggum Takes Over Coding, and GPT Health Arrives Listen now | From Weights & Biases, latest ThursdAI roundup with NVIDIA CES news, Grok no guardrails, Ralph Wiggum breakdown with Ryan, GPT Health and OSS AI!

Also ChatGPT Health launched (partners with Color Health for prescriptions), @LTXStudio LTX-2 went fully open source for video generation, and AI doctors are now legally prescribing in Utah

We are speedrunning the future and I'm here for it

Full notes: thursdai.news/jan-8

.@ryancarson joined us to break down the Ralph Wiggum technique that went viral

Basically: vibe coding is dead, structured outputs + smaller models are beating raw prompting at scale, and we're seeing the shift from "just throw more tokens at it" to actual engineering

This is huge

NVIDIA just announced Vera Rubin at CES and it's absolutely wild

50 petaFLOPS of FP4 performance, 5x faster than Blackwell, and they're shipping THIS YEAR

Meanwhile XAI raised $20B to build the world's largest supercomputer in Memphis. The compute race is getting insane

You can read the full notes here thursdai.news/jan-8 and see the edited video on YT here:

youtu.be/_8zf5BtF-Ts

Colorful sketchnote summarizing tech headlines: NVIDIA Vera Rubin GPU leap, open‑source model updates, AI health tools, Gmail Gemini, voice AI, plus a cartoon host.

🎙️ First live @thursdai_pod of 2026 is UP and holy shit what a way to start the year

NVIDIA drops 5x performance leap at CES, XAI raises $20B (while causing bikini-gate), ChatGPT officially goes into Health, and Ryan Carson breaks down why Ralph Wiggum is the death of vibe coding

Let's goooo 🧵

Full episode is also on YouTube if you want the deep dive with quarterly breakdowns and all the context:

youtu.be/F1RtlhyErmg

Happy holidays everyone! 🎅 Signing off till next year 🫡

See you in 2026 when things get even more wild

Massive shoutout to the cohosts who made this year possible 🙏

@WolframRvnwlf @yampeleg @nisten @ldjconfirmed @ryancarson @kwindla

you all rock and made covering this insane year actually fun instead of just overwhelming