Pioneering STARKs since 2018 for blockchain scale, privacy, and post-quantum security.
Joined January 2018
- Tweets13.3K
- Following452
- Followers232K
- Likes18.8K
Pinned Tweet
Anthropic is using cryptography StarkWare’s founders co-invented to prove which AI model answered you.
With FlashProve, you can verify that a specific model produced an output while keeping its weights private.
The report cites the original STARK paper, highlights the use of STARKs on @Starknet, and references formal verification work on Cairo and the S-two AIR. StarkWare co-founder Alessandro Chiesa is one of the report’s co-authors.
Another big validation for our Proving System.
The same cryptographic ideas that help scale blockchains are now helping make AI inference verifiable.
StarkWare retweeted
STARKs are finding applications far beyond blockchain scaling.
Anthropic's FlashProve shows how the cryptography behind Starknet can help verify AI model outputs without exposing private weights.
Different use case. Same cryptographic foundation.
Starknet is part of a much bigger story.
Anthropic is using cryptography StarkWare’s founders co-invented to prove which AI model answered you.
With FlashProve, you can verify that a specific model produced an output while keeping its weights private.
The report cites the original STARK paper, highlights the use of STARKs on @Starknet, and references formal verification work on Cairo and the S-two AIR. StarkWare co-founder Alessandro Chiesa is one of the report’s co-authors.
Another big validation for our Proving System.
The same cryptographic ideas that help scale blockchains are now helping make AI inference verifiable.
StarkWare retweeted
Fuck it I'm frontrunning @LootRealms announcement
300,000 $LORDS are up for grabs at @EFDevcon !
Realms: Frontier's S0 is live for 4 days.
Every day a new map rises out of the mist. Send your armies in, beat what guards it, and drag $LORDS out of the ruins and chests you find.
In 2018, our founders invented STARKs. We then put them into production to scale blockchains and bring privacy onchain.
Now Anthropic is building on that same cryptography to make AI inference verifiable.
Love seeing STARK research find new applications beyond blockchains.
New Anthropic research: progress on provable inference for large language models.
Here, we introduce FlashProve. Building on established cryptographic techniques, FlashProve allows developers to verify whether a given response came from a given AI model. It’s substantially more efficient than prior open-source methods for doing so.
These proofs have a range of applications. They can, for instance, make it harder for attackers to hijack a chat, or enable safety evaluators to verify that the model they’re testing is the same one being served to customers.
For our full write-up, see here: anthropic.com/research/prova…
Further progress in provable inference will likely be an important part of improving trust in AI deployment. To that end, we're open-sourcing FlashProve’s code, here: github.com/anthropics/FlashP…
We’re grateful to our collaborators: Alessandro Chiesa, @GiacomoFenzi, @gwberg95, @miranda__christ, Rahul Dalal (rdalal.github.io) and Ben Mayer (@transmissions11).
StarkWare retweeted
Now imagine what Starknet L1 will look like!
Prove in the most efficient manner everything that has value:
- Private Assets (now)
- Inference
- Proof of Reserves
- Interchain / Interbank txs
- IP / Identity / Ownership
Anthropic is using cryptography StarkWare’s founders co-invented to prove which AI model answered you.
With FlashProve, you can verify that a specific model produced an output while keeping its weights private.
The report cites the original STARK paper, highlights the use of STARKs on @Starknet, and references formal verification work on Cairo and the S-two AIR. StarkWare co-founder Alessandro Chiesa is one of the report’s co-authors.
Another big validation for our Proving System.
The same cryptographic ideas that help scale blockchains are now helping make AI inference verifiable.
StarkWare retweeted
STARKs enforce integrity, and in this case, for LLMs.
I'm amazed by how far we've come: from "We don't believe STARKs can do what you say they can" in 2018, to STARKs can deliver scale, privacy, post-quantum security, and now also enforce integrity for LLMs.
Anthropic is using cryptography StarkWare’s founders co-invented to prove which AI model answered you.
With FlashProve, you can verify that a specific model produced an output while keeping its weights private.
The report cites the original STARK paper, highlights the use of STARKs on @Starknet, and references formal verification work on Cairo and the S-two AIR. StarkWare co-founder Alessandro Chiesa is one of the report’s co-authors.
Another big validation for our Proving System.
The same cryptographic ideas that help scale blockchains are now helping make AI inference verifiable.
StarkWare retweeted
The MOB is LIVE! ⚔️
7,068 Musical Onchain Beasts await a worthy Adventurer. 👑
🎮 Game by @ProvableGames
🎼 Music by @CaseyWescott
🎨 Art by @1337skulls
🔋 Powered by @Starknet
Play to Collect 👇
lootsurvivor.io/beastmode
You know what your favorite collectible looks like. 🔍
Now imagine discovering what it sounds like. 🎵
That’s the next chapter of the Beasts. 🧌
🔊 Sound On 🔊
@ProvableGames @CaseyWescott @1337skulls @Starknet
StarkWare retweeted
anthropic just proved what a 400b model said
and their own report says the proofs are STARKs
FlashProve, from anthropic's red team:
> a lab commits to its model weights once
> every reply ships with a ~19mb proof
> your computer checks it in ~17 seconds, no gpu
> swap the model and the check fails
> the weights stay hidden
STARKs
> only hash functions
> no elliptic curves
> post quantum
the report also:
> calls STARKs "battle-tested" and names starknet as an example
> has alessandro chiesa (StarkWare co-founder) as an author
> starts its recipe from a 2016 paper by eli ben-sasson
starkware put the first STARKs on ethereum mainnet in 2020
this is a STARK win
StarkWare retweeted
🚨BREAKING : ANTHROPIC JUST MENTIONED STARKNET IN THEIR TECHNICAL REPORT
anthropic just published how to prove an LLM output came from the exact model claimed using a hash based zk proofs that are post-quantum secure.
starknet got cited as an example of a chain in production that does the same.
this makes it the 2nd company after @GoogleQuantumAI which mention about starknet, Starkware, zkstarks that’s live post quantum secure examples.
StarkWare retweeted
Starknet is an AI chain.
🚨BREAKING : ANTHROPIC JUST MENTIONED STARKNET IN THEIR TECHNICAL REPORT
anthropic just published how to prove an LLM output came from the exact model claimed using a hash based zk proofs that are post-quantum secure.
starknet got cited as an example of a chain in production that does the same.
this makes it the 2nd company after @GoogleQuantumAI which mention about starknet, Starkware, zkstarks that’s live post quantum secure examples.
StarkWare retweeted
→ March 31, 2026: Google Quantum AI points to Starknet as a live example of hash-based proofs believed to resist quantum attacks.
→ October 7, 2026: Europol discusses zk-STARKs as a tool for reducing wallets’ exposure to quantum attacks and cites StarkWare in its report.
→ November 2, 2026: Anthropic releases FlashProve for verifiable AI inference, points to Starknet as a deployment of the same kind of hash-based proofs, and credits StarkWare researcher as a contributor.
Love seeing the cryptography StarkWare pioneered show up in everything from post-quantum security to verifiable AI.
If you’d rather read it on StarkWare’s blog, head over here 👇
starkware.co/blog/why-quantu…
Programmable privacy at planetary scale, built for a post-quantum world.
Starknet is the endgame.
Hear it directly from @EliBenSasson.
11/ Higher².
teapawt.net/SuperVegaXYZ/status/21…
October has been our best month by far at @SuperVegaXYZ 🔥
Glad to become the home of more and more option traders everyday
Onward -> Better UI, more assets, more options
12/ Note: this is NFA and is provided for informational purposes only.
Stay SAFU and always DYOR.
For a complete recap of last week’s news, check it out here:
starknetresearch.substack.co…