Tom V

Most online sources state that the earliest use of computer vision is "The Summer Vision Project" in 1966 at MIT [1] but there is an amazing project that predated it that deserves way more credit - It involves corn in Indiana and the invention of MLOps🧵 (1/12)

@siddarthv66 Self-driving is robotics, clear that the AGI companies will not solve self-driving before the incumbents.

Robotics will be solved by AGI companies (Ant/OAI/GDM) before robotics companies (PI, Figure, Skild)

@chris_j_paxton Great summary - would add @rhodaai as another video world model with inverse dynamics and @wayve_ai's GAIA as an action-conditioned world model.

A new blog post on world models: - what do people mean when they say world models? - why are they exciting? - what might allow them to solve a wide range of robotics problems?

@aakashgupta "Google spent a decade getting mocked for falling behind OpenAI" - more accurate would be 2022-2024 and anyone serious in the industry knew Google would be able to deliver sota models once they prioritized scaling.

Meta is about to spend $135 billion in capex this year to license someone else’s AI. Zuckerberg made the call himself. Llama 4 flopped in April 2025. Instead of fixing the team he had, he paid $14.3 billion to poach Scale AI’s Alexandr Wang, blew up the entire AI org, created Meta Superintelligence Labs, recruited the former GitHub CEO, hired a co-creator of ChatGPT, and imposed 70-hour workweeks on a company that used to run on consensus and committee. The man who mass-fired 21,000 employees during the “Year of Efficiency” decided the problem was he hadn’t spent enough money. Eleven months and billions later: Avocado underperformed Google’s Gemini 3.0 on internal benchmarks and just got delayed to May. That’s two consecutive flagship model failures in 12 months. Now Meta is reportedly considering licensing Google Gemini to power Meta AI while Avocado bakes longer. The same Google that just signed a $1 billion per year deal to run Apple’s Siri. The same Google whose Gemini models are now the intelligence layer behind 1.5 billion iPhones. Run the math on what Google is assembling. Apple: 1.5 billion devices. Meta: 3.6 billion MAUs across Facebook, Instagram, and WhatsApp. If both deals close, Google’s AI models would sit behind roughly 5 billion user touchpoints. No other company is close. Google spent a decade getting mocked for falling behind OpenAI. While everyone was writing the obituary, Pichai was building the infrastructure that makes Gemini the enterprise default. Apple evaluated OpenAI, Anthropic, and Google. Google won on performance AND price. Meta’s 2026 capex guidance is $115 to $135 billion. The company spending more on AI infrastructure than all but 50 countries’ GDPs might end up routing its 3.6 billion users through a competitor’s model. The distribution moat everyone assumed Meta had was always the apps, never the models. Google just proved it.

@fhuszar Well done Kseniia!! 💙💛

Kseniia, our youngest scholar was only 12yo when she wrote in 2022. After lots of programming competition medals and her first IOI last year: She scored a gold at the Int’l AI Olympiad, having worked on a diffusion LM side project since participating in airetreat.org

@ollieforsyth Spain has higher wealth taxes than the UK there is the "Beckham" law but that's only for a finite number of years. People are leaving to Spain to get more sun in the age of remote work not because of tax.

The UK has everything going for it when it comes to founding: Talent, Capital, Community, Networks, Universities, Global Travel Hub and so on. But if your own government doesn't incentivize and celebrate amazing founders and companies and instead, taxes them to the bone after years of sleeping in the office, THEY WILL LEAVE. Simple. Source @FT: t.co/MWFdgiHvnk

Has any major B2B SaaS enabled a "Generate Feature" button where instead of a support ticket you can prompt a custom frontend view of the platform for your needs?

@SimEdw Try a log plot! I see some signal

75M → good 30M → same 9M → ...almost same?? Model scaling laws taking a vacation 😂

@BillLeaver_ @GammaApp Just did, not a fan, it needs to be deeply integrated into the slide editor and it needs to be able to rearrange elements using artistic direction.

@tomvarsavsky Tried @GammaApp?

Contained within the weights of Nano Banana is an understanding of visual aesthetics and contained within the weights of Claude opus 4.5 is intelligent tool use - can someone please train a frontier model that can actually make a good slide deck?

Instead of buying out top researchers from each other AI labs should fund postdocs, PhDs and other initiatives like math/cs Olympiads. They would grow the pie and stop making it a zero sum game. It's quite short sighted, and makes "racing china" arguments hard to defend.

The most worthwhile unsolved but solvable ML problems in industry are in biochemistry, materials science, physical AI and foundation models (geometric, video, audio, LLMs, reasoning). You can work on any of these in the UK and avoid "fix recruitment" or "automate BDR" companies.

@EastlondonDev The best teachers find a way to make you feel good about your questions, we're already being coddled by superior minds

The most flattering AI will win the consumer market The same way that people prefer optimistic weather forecasters over accurate ones

@EastlondonDev Well it's android for me - intercepting them would be a good way to give the app eyes on to all my incoming requests/messages without authenticating everywhere. I could put my apps on more verbose settings and delegate the filtering to this notifications agent.

@tomvarsavsky Ah so you want something that intercepts Apple push notifications ?

I got a wake up phone call from my AI via MCP!

@EastlondonDev Well currently it's my phone's notification service but something that might work well is just a WhatsApp account which always beeps when it reaches out.

@tomvarsavsky Nice, How do your apps send you notifications? Phone #, email, other?

@EastlondonDev I want either an android app or an offline server that can reach me that I give maximum privileges to and receives all the notifications from my apps and decides what needs my attention based on a conversation. E.g "if xxxx reaches me on any app make the loudest noise possible"

@EastlondonDev Are you aware of any cool "LLM managed phone notifications" projects? Being able to configure your notification profiles across apps using natural language

Leaving this as a prediction for the future - We've tapped out on exam style evals, next stage is agent evals. These will initially be in controlled environments and eventually be measured simply in the revenue that they can generate in the real world.

@karpathy any plans to do education programs for K-12 through @EurekaLabsAI ?

@azeem 🎯

🌞 10 weeks of releases in 1 update. What we chose to build and why 🧵