Max Sapo @MaxHappyverse

If you've wondered where I've been recently, I gave an interview. This is a very difficult time for me. I'm trying to bring my mom home, and I could use your help spreading the word. Please watch, and please share. interview: youtube.com/watch?v=JlK2a6… website: bringmymomhome.com

Successfully training models on TPUs has been demonstrated by Anthropic through the past five-plus successful Claude releases. This is positive for the ML community, as Google TPUs continue to gain market share outside of internal Google workloads, giving frontier AI labs a viable alternative for training. 1/4🧵

Reading what AI doesn't say (this topic genuinely matters to me, and it happens to be my birthday today, so if you feel like it, a repost would be a nice gift 🎂) Something shifting in AI architecture that I don't think gets discussed enough through a safety lens. Models are starting to pass information to each other directly through internal representations, hidden states, activations, and computation vectors (still mostly research right now, but the infrastructure will follow). That matters for how we audit reasoning, since most of what we can currently check comes from what shows up in words. Right now, the main tool we have for auditing AI reasoning is reading chain-of-thought. Korbak et al. (2025) spend a whole paper arguing this window is already "fragile" for single-model reasoning. If models start reasoning through vectors passed model-to-model instead of text, understanding that latent layer becomes the only way to keep any visibility into what's actually happening. Thus I think that what seems underexplored is understanding the math of these latent spaces, and it might be the same research direction as learning how to monitor them. Models appear to converge toward similar geometric structure regardless of architecture (Huh et al. 2024). If that's right, probes for safety-relevant features might generalize: deception patterns, goal representations, misalignment signals studied once and applied across models rather than re-derived per architecture. Whether this holds for safety-critical features specifically is open, so is what adversarially robust latent decoding would even look like. Both feel more urgent than the current research investment suggests.

Congrats @spurs 🤙

@sksuvorovas do you also get mail from California state saying you can sign up for food stamps because they found that your founder income is too low so you can qualify?🙃

@sksuvorovas Yeah pretty good for deck ideas and for taking to a bunch of pdfs

🚀Today we ship @FlyMy_AI Agents. The world's first all-in-one agentic cloud. The modern way to build, integrate, and scale production AI agents. 3 steps to a production agent: 1. Connect your work tools to FlyMy 2. Describe what the agent should do - in text or 5 lines of code 3. Set execution rules: manual, scheduled, or integrated into your backend 4. Done! Agent works on scale! Everything in one place: 800+ MCPs, hundreds of AI models, brain, memory, sandboxes. Stop building from scratch. Stop waiting for infra. Compress 6 months into a day. #1 on @ArtificialAnlys benchmarks. Stable, secure, scalable from day one. Try FlyMy.AI →

Amin Vahdat is one of the most underrated leaders of our generation. His ability to zoom in / zoom out is just second-to-none.

Shanu Mathew @ShanuMathew93

3 months ago

Amin Vahdat live at Transition-AI 2026 with @CatalystPod. Google's Chief Technologist for AI Infrastructure - the man in charge of Google's $175–185B 2026 CapEx and the one who's said compute capacity needs to 2x every 6 months! Key takeaways: Training → Inference cascade:

keyboard_arrow_left Previous keyboard_arrow_right Next

6 17 172 43K 268

@sundeep 1M+ TPU 8t though, slightly more than 960K A5Xs :)

@gdb Cool, just sent an email :)

@sama Is Codex on Cerebras and your own ASICs coming soon as well?

@JeffDean @AcquiredFM @gilbert @djrosent It was such a great session @JeffDean - you and Amin are an amazing duo - thanks for doing so many great things for the world! excited to try out TPU v8 when it becomes available!

@DynamicWebPaige @GoogleDeepMind @GoogleStartups @EnchantedTools @GoogleAIStudio @livekit We like PipeCat a little more :)

Wow, even @elonmusk acknowledges this! #GoTPUs!

Elon Musk @elonmusk

3 months ago

@sundarpichai TPUs are underrated

251 319 6K 429K 235

cloud.google.com/blog/products/…

Met lots of good old friends at @googlecloud #Next2026 and even took a selfie with #TPUv4, my first product launch that I led back in 2022. Can’t believe it’s been 4 years ago - kids grow fast :) Link to a blog in comments below (for those curious to compare what was state-of-the-art then vs now. 🙂)

keyboard_arrow_left Previous keyboard_arrow_right Next

@Jason I’m in

Met @que_tourist earlier this week. My best celebrity selfie of the year so far 🤗🤙

@HappyverseAI webinar on 2/26! Tomorrow we are teaching a diverse group of founders, knowledge workers and creators how to build and deploy their own Digital Employee — a real-time AI video agent that can sell, screen, coach, and represent you 24/7. Day 5 of our AI Bootcamp covers: → The architecture behind sub-500ms real-time video agents → Writing system prompts that actually govern behavior → Connecting your agent to calendars, Google Meet/Zoom and agentic tools → Security, guardrails, and going from demo to business asset You leave with a working, deployable digital employee — not a concept. Joining me are my colleagues and advisers - Nora Salim and Mariana Tataryn. Last chance to grab a seat: luma.com/s4uydmv4 Use code HAPPY for a discount. See you there 🤝

@dylan522p Go with Inference Dylan