abdel

@AbdelStark • 31,039 subscribers

Head of Applied AI & Verifiable Intelligence @StarkWareLtd. STARK proofs, AI Safety, World Models & Physical AI / Robotics. Math scale, Goodwill doesn't.

Shorts

Can LLMs be PROVABLE computers? Percepta showed that a transformer can BE a computer. Compiled weights, deterministic execution, 30k tokens/sec. But nobody asked the obvious follow-up: how do you know it computed correctly? So I built the verification layer. A STARK that proves it 👇

Can LLMs be PROVABLE computers? Percepta showed that a transformer can BE a computer. Compiled weights, deterministic execution, 30k tokens/sec. But nobody asked the obvious follow-up: how do you know it computed correctly? So I built the verification layer. A STARK that proves it 👇

183,163 views

bUT WhAt CaN YoU Do WiTh ThoSE PRoOfs ? Here you go: TrustMeBro is a showcase of a mini ZK-ready Bitcoin explorer where anyone can verify transaction and block validity right in their browser, instant, trustless, no middlemen. Think about it like TLS. 99.99% of people browsing the web have no clue how a secure session actually works, but they've all internalised that green lock meaning "you're safe." ZK can be the same. Run a local verification under the hood, flash a simple badge, verified or not, and boom! Starknet tech looking good from here. As I said many times, Bitcoin x STARKs is a math match made in heaven. Won't stop till the whole damn world is proven. One STARK at a time.

bUT WhAt CaN YoU Do WiTh ThoSE PRoOfs ? Here you go: TrustMeBro is a showcase of a mini ZK-ready Bitcoin explorer where anyone can verify transaction and block validity right in their browser, instant, trustless, no middlemen. Think about it like TLS. 99.99% of people browsing the web have no clue how a secure session actually works, but they've all internalised that green lock meaning "you're safe." ZK can be the same. Run a local verification under the hood, flash a simple badge, verified or not, and boom! Starknet tech looking good from here. As I said many times, Bitcoin x STARKs is a math match made in heaven. Won't stop till the whole damn world is proven. One STARK at a time.

32,110 views

Videos

Anya Rossi

sweetdream.ai

SweetDream.ai•Sponsored•Livecam

Watch Anya Live

Anya is streaming live right now! Join her private show and enjoy exclusive content.

Exclusive private shows

1.2k viewers online

Private Show

Join now for exclusive access

Free preview available • Premium content

Congrats to the Kimi.ai team! This is awesome. Great to see this level of research coming from open-source frontier model labs. I liked the paper so much I built a Rust implementation of it ;) Full AttnRes + Block AttnRes with two-phase inference, built using Burn (tensor library and Deep Learning Framework, in Rust, by Tracel AI). Runs on CPU, CUDA, Metal, wgpu. Includes an interactive TUI that trains a model live and visualizes depth attention evolving from uniform to selective in real time. Repo link and more on what is implemented in the comments.

Congrats to the Kimi.ai team! This is awesome. Great to see this level of research coming from open-source frontier model labs. I liked the paper so much I built a Rust implementation of it ;) Full AttnRes + Block AttnRes with two-phase inference, built using Burn (tensor library and Deep Learning Framework, in Rust, by Tracel AI). Runs on CPU, CUDA, Metal, wgpu. Includes an interactive TUI that trains a model live and visualizes depth attention evolving from uniform to selective in real time. Repo link and more on what is implemented in the comments.

95,651 views • 5 months ago

Why LLMs are a dead end for human-level intelligence, and especially for Physical AI / Robotics. The next leap isn’t bigger language models. It’s World Models. I just dropped a full 1-hour presentation from Shanghai: “World Models: the ChatGPT moment for robotics?” → Why LLMs hit a wall → Why action-conditioned world models planning in latent space are the real path → Live World Forge demo with LeWorldModel + Hugging Face LeRobot Watch here. The future of intelligence is embodied, not just chatty.

Why LLMs are a dead end for human-level intelligence, and especially for Physical AI / Robotics. The next leap isn’t bigger language models. It’s World Models. I just dropped a full 1-hour presentation from Shanghai: “World Models: the ChatGPT moment for robotics?” → Why LLMs hit a wall → Why action-conditioned world models planning in latent space are the real path → Live World Forge demo with LeWorldModel + Hugging Face LeRobot Watch here. The future of intelligence is embodied, not just chatty.

38,919 views • 3 months ago

I am stocked to announce that I won the OpenAI Developers Codex x Mollie Hacka Worldwide Hackathon in Paris. 60+ builders, every one of us working solo, one day to ship. I built mine around a single question: who gets to own intelligence? The default answer is scary. You hand your data to a handful of labs, they train the model, they own it, and you rent back a thin slice of what your own data made possible. That is the bargain on the table today. I do not accept it. So I built Lensemble: a Tapestry like distributed training platform for JEPA based World Models. What does it enable: World Models that a community improves together, keeps sovereign, and co-owns. Two bets sit underneath it. First, the paradigm. Language models predict the next token. Powerful for text, a dead end for the physical world. A robot does not need to autocomplete sentences, it needs to predict what happens next in the world. That is what JEPA does: it learns by predicting representations instead of pixels or tokens. I am convinced world models are the most underrated paradigm in AI right now, and the closest thing we have to a ChatGPT moment for robotics. Second, the politics. Your raw trajectories never leave your machine. Each participant trains locally against a shared protocol and ships only an update, never the data. A federated round folds those updates into one shared world model, a LeWorldModel based model, and the gain is measured, not claimed: a 12k-parameter adapter on a frozen backbone, held-out prediction error down about 12 percent, the model measurably less surprised by the world. Then the upside is split by contribution weight, so the people who improved the model own a share of what it earns. This is the thesis behind Project Tapestry, the AI Alliance and Yann LeCun's push for federated, sovereign frontier AI, carried into world models and robotics. Call it Tapestry for the physical world. All of it built solo, in a single day, with Codex as my pair the whole way. Thank you to OpenAI Codex and Mollie for backing builders who ship real things, and to Boris and the organizing crew for the room and the standard you set. Intelligence the world improves, and the world owns. That is the future I want for my kids, and the one I will keep building.

I am stocked to announce that I won the OpenAI Developers Codex x Mollie Hacka Worldwide Hackathon in Paris. 60+ builders, every one of us working solo, one day to ship. I built mine around a single question: who gets to own intelligence? The default answer is scary. You hand your data to a handful of labs, they train the model, they own it, and you rent back a thin slice of what your own data made possible. That is the bargain on the table today. I do not accept it. So I built Lensemble: a Tapestry like distributed training platform for JEPA based World Models. What does it enable: World Models that a community improves together, keeps sovereign, and co-owns. Two bets sit underneath it. First, the paradigm. Language models predict the next token. Powerful for text, a dead end for the physical world. A robot does not need to autocomplete sentences, it needs to predict what happens next in the world. That is what JEPA does: it learns by predicting representations instead of pixels or tokens. I am convinced world models are the most underrated paradigm in AI right now, and the closest thing we have to a ChatGPT moment for robotics. Second, the politics. Your raw trajectories never leave your machine. Each participant trains locally against a shared protocol and ships only an update, never the data. A federated round folds those updates into one shared world model, a LeWorldModel based model, and the gain is measured, not claimed: a 12k-parameter adapter on a frozen backbone, held-out prediction error down about 12 percent, the model measurably less surprised by the world. Then the upside is split by contribution weight, so the people who improved the model own a share of what it earns. This is the thesis behind Project Tapestry, the AI Alliance and Yann LeCun's push for federated, sovereign frontier AI, carried into world models and robotics. Call it Tapestry for the physical world. All of it built solo, in a single day, with Codex as my pair the whole way. Thank you to OpenAI Codex and Mollie for backing builders who ship real things, and to Boris and the organizing crew for the room and the standard you set. Intelligence the world improves, and the world owns. That is the future I want for my kids, and the one I will keep building.

20,169 views • 2 months ago

Here is a full demo of an onchain verification of a STARK proof of a Fibonnacci computation, on Catnet (custom Bitcoin signet with OP_CAT enabled). Based on the amazing work of !

Here is a full demo of an onchain verification of a STARK proof of a Fibonnacci computation, on Catnet (custom Bitcoin signet with OP_CAT enabled). Based on the amazing work of !

76,649 views • 1 year ago

Provable SHRIMPS: Post-quantum hash-based signatures verified in Cairo with STARK proofs. As usual Blockstream Research is killing it on the post quantum bitcoin research. SHRIMPS 🦐 is so cool that I made provable shrimps, it could be useful to for batch shrimps signature verification. Don't panic, as Adam Back said: "SHRIMPS & bitcoin quantum readiness > quantum FUD."

Provable SHRIMPS: Post-quantum hash-based signatures verified in Cairo with STARK proofs. As usual Blockstream Research is killing it on the post quantum bitcoin research. SHRIMPS 🦐 is so cool that I made provable shrimps, it could be useful to for batch shrimps signature verification. Don't panic, as Adam Back said: "SHRIMPS & bitcoin quantum readiness > quantum FUD."

25,494 views • 5 months ago

Introducing ProvableWorldModel: the first provable JEPA world model. A real pretrained world model runs, and anyone can verify, on a laptop CPU, or on a mobile phone, that the committed model ran exactly as claimed, in milliseconds. The whole industry spent this week asking one question about Anthropic's Fable 5: which model actually answered me? Nobody could prove it. Not during the silent downgrades, not after the apology. The fix is a notification you still have to trust. I've been building the alternative. The intuition is simple. World models are small. Small enough to run in exact integer arithmetic, where every operation is reproducible bit for bit, attention included. The prover commits to its full execution trace before any challenge is derived. The verifier fingerprints the heavy matmuls with 1979-era randomized algebra (Freivalds) and replays everything else exactly. Change one number anywhere and the proof dies. Receipts: → real pretrained JEPA predictor (LeWorldModel), 10 MiB int8, real expert episode → inference 49 ms, verification 29 ms, CPU only, no floats, no GPU → soundness error ≈ 2^-44 → forge a single matmul output and the verifier throws FreivaldsCheckFailed Why it matters: LLM agents need ever-thicker invisible harnesses to behave, and harnesses can't be audited from the outside. World models invert that. Small, structured, exact: verifiable by construction. Intelligence you can check, not intelligence you have to trust. Don't trust, verify.

Introducing ProvableWorldModel: the first provable JEPA world model. A real pretrained world model runs, and anyone can verify, on a laptop CPU, or on a mobile phone, that the committed model ran exactly as claimed, in milliseconds. The whole industry spent this week asking one question about Anthropic's Fable 5: which model actually answered me? Nobody could prove it. Not during the silent downgrades, not after the apology. The fix is a notification you still have to trust. I've been building the alternative. The intuition is simple. World models are small. Small enough to run in exact integer arithmetic, where every operation is reproducible bit for bit, attention included. The prover commits to its full execution trace before any challenge is derived. The verifier fingerprints the heavy matmuls with 1979-era randomized algebra (Freivalds) and replays everything else exactly. Change one number anywhere and the proof dies. Receipts: → real pretrained JEPA predictor (LeWorldModel), 10 MiB int8, real expert episode → inference 49 ms, verification 29 ms, CPU only, no floats, no GPU → soundness error ≈ 2^-44 → forge a single matmul output and the verifier throws FreivaldsCheckFailed Why it matters: LLM agents need ever-thicker invisible harnesses to behave, and harnesses can't be audited from the outside. World models invert that. Small, structured, exact: verifiable by construction. Intelligence you can check, not intelligence you have to trust. Don't trust, verify.

15,377 views • 2 months ago

Introducing WorldForge: testable world-model workflows for physical AI systems. You can think of it, loosely, as “LangChain for world models”. The problem is that “world model” has become an overloaded label. Depending on context, it can mean a video generator, a cost model, a robot policy, a JEPA-style latent predictor, etc. They share almost nothing, different inputs, runtimes, failure modes. I built WorldForge to stop pretending they're interchangeable. Front-door demo: a real Hugging Face LeRobot (LeRobot) diffusion_pusht policy combined with LeWorldModel by Lucas Maes checkpoint for scoring. Both run locally on my MacBook in the demo video. LeWM is extremely efficient (~15M params), can plan up to 48× faster, and runs on commodity hardware. WorldForge wires the loop: policy → candidates → score → select Replay happens in a local TUI today, but the same loop could drive a real robot in the world. Would love feedback from people working on world models, physical AI, robotics, ML infra, and adjacent tooling. Fully open source. Contributions very welcome. Plan in the dream, replay in real world.

Introducing WorldForge: testable world-model workflows for physical AI systems. You can think of it, loosely, as “LangChain for world models”. The problem is that “world model” has become an overloaded label. Depending on context, it can mean a video generator, a cost model, a robot policy, a JEPA-style latent predictor, etc. They share almost nothing, different inputs, runtimes, failure modes. I built WorldForge to stop pretending they're interchangeable. Front-door demo: a real Hugging Face LeRobot (LeRobot) diffusion_pusht policy combined with LeWorldModel by Lucas Maes checkpoint for scoring. Both run locally on my MacBook in the demo video. LeWM is extremely efficient (~15M params), can plan up to 48× faster, and runs on commodity hardware. WorldForge wires the loop: policy → candidates → score → select Replay happens in a local TUI today, but the same loop could drive a real robot in the world. Would love feedback from people working on world models, physical AI, robotics, ML infra, and adjacent tooling. Fully open source. Contributions very welcome. Plan in the dream, replay in real world.

17,001 views • 4 months ago

Breaking news coming from El Salvador today, let me introduce you the first Starknet powered Bitcoin ATM in the world! Edgar Borja did the first ever Bitcoin purchase from a K1 Technology ATM where the funding is coming from a Starknet wallet via an atomic swap to Lightning!

Breaking news coming from El Salvador today, let me introduce you the first Starknet powered Bitcoin ATM in the world! Edgar Borja did the first ever Bitcoin purchase from a K1 Technology ATM where the funding is coming from a Starknet wallet via an atomic swap to Lightning!

28,977 views • 1 year ago

L402 is HTTP for payments, literally, and it's cool! Built a Rust implementation for an API paywalled with the L402 protocol, as well as a demo frontend. Try it yourself: Pay with Lightning and query the latest Bitcoin block.

L402 is HTTP for payments, literally, and it's cool! Built a Rust implementation for an API paywalled with the L402 protocol, as well as a demo frontend. Try it yourself: Pay with Lightning and query the latest Bitcoin block.

18,073 views • 1 year ago

Wanna start dev Starknet projects ? Made a quick tool using kickstart, a Rust CLI to easily get a new project up and running by using pre-made templates. It will create a ready to use project using the great Starknet Foundry toolchain.

Wanna start dev Starknet projects ? Made a quick tool using kickstart, a Rust CLI to easily get a new project up and running by using pre-made templates. It will create a ready to use project using the great Starknet Foundry toolchain.

13,906 views • 3 years ago

No more content to load