Tomiwa Adey

Building and experimenting with 3D Agents

[Tomiwa Adey @tomiwaAdey]

In tech, $1 million in investment gets you tons of tech press, but $1 million in revenue no one cares about. 🔥

[Tomiwa Adey @tomiwaAdey]

@wawasensei 98% AI 😂

[Tomiwa Adey @tomiwaAdey]

@recap_david Calico

[Neil Zeghidour @neilzegh]

Me defending my O(n^3) solution to the coding interviewer.

[Tomiwa Adey @tomiwaAdey]

@bnj @variantui I’d love to try - eng

[Tomiwa Adey @tomiwaAdey]

@skirano 👋🏾

[Tomiwa Adey @tomiwaAdey]

@MEonEVM Gm

[Balaji @balajis]

It might turn out that the GPT wrapper has more of a moat than GPT.

[Tomiwa Adey @tomiwaAdey]

@nateberkopec Yup, however you can start simple e.g read replicas + materialised views is a powerful but simple start.

[Tomiwa Adey @tomiwaAdey]

@Thom_Wolf The focus on individuals is overblown.

[Tomiwa Adey @tomiwaAdey]

@benln Building 👋🏾

[Tomiwa Adey @tomiwaAdey]

@emaalozada @v0 @AdventOfPrompt Seriously how did you do this?

[Tomiwa Adey @tomiwaAdey]

@LuLing26466911 @zhou_xian_ He confirmed it in this tweet
https://t.co/CAOzr9QZCa
"and a upper level LLM/VLM agent that uses the physics engine as tools to compose worlds and generate other modalities of data."

I'm working on exactly this so I was able to guess correctly.

[Zhou Xian @zhou_xian_]

Hi Jon, thank you for your interest! The project is essentially two parts: a low level physics engine that integrates a wide range of solvers for simulating materials with different physical properties, and a upper level LLM/VLM agent that uses the physics engine as tools to compose worlds and generate other modalities of data. The generative agent incorporates multiple generative modules handling different modalities in a modular way, and the high-level goal is instead of generating everything end-to-end, we generate them in a structured way and all of them collectively form a physical world :) Let me know if you want to dive deeper

[François Chollet @fchollet]

Today OpenAI announced o3, its next-gen reasoning model. We've worked with OpenAI to test it on ARC-AGI, and we believe it represents a significant breakthrough in getting AI to adapt to novel tasks.

It scores 75.7% on the semi-private eval in low-compute mode (for $20 per task in compute ) and 87.5% in high-compute mode (thousands of $ per task). It's very expensive, but it's not just brute -- these capabilities are new territory and they demand serious scientific attention.

[Tanishq Mathew Abraham, Ph.D. @iScienceLuvr]

"o3-mini by end of Jan, full o3 shortly after that"

- Sam Altman on public release of o3 models

[Tomiwa Adey @tomiwaAdey]

@LuLing26466911 @zhou_xian_ @zhou_xian_ ?

[Tomiwa Adey @tomiwaAdey]

@LuLing26466911 @zhou_xian_ Probably not. I don’t think they use diffusion-based priors at all. It’s likely an LLM based generation with visual eval.

[Tomiwa Adey @tomiwaAdey]

@simonw I was waiting for you to do this 😂 I think people are sleeping on this feature.

[Tomiwa Adey @tomiwaAdey]

@daniel_levine Yes please 🙏🏾

[Tomiwa Adey @tomiwaAdey]

@EHuanglu I’m working on this.