Independent Web3 researcher Testnets • Verification • Protocol analysis Don’t trust. Verify.
Ukraine
- Tweets 3,076
- Following 1,620
- Followers 232
- Likes 3,930
Pinned Tweet
173+ days later.
Full assessment:
github.com/Sergey007S/psy-pr…
The core technical question remains unchanged:
Can the reported 521k TPS benchmark be independently reproduced from the public artifacts available for verification?
No public reproducibility review.
No public benchmark reproduction.
No public resolution of the challenge.
cc: @cmpeq @cyberqualia @StarkWareLtd @bcap
@PsyProtocol
1/5
March 4, 2026.
More than 100 days ago, I started trying to independently verify a benchmark that Psy Protocol publicly promoted as:
• 521k TPS
• “verifiable by anyone”
• backed by a $100,000 bounty
The problem was simple:
Despite the “verifiable by anyone” claim, I could not find a practical public way for an independent researcher to verify whether the benchmark was actually representative.
Over the next 100+ days I:
• published public questions
• performed technical analysis
• contacted the team directly
• submitted a formal bounty claim
• reached out to investors connected to the project
The core issue remains unresolved.
No practical public verification path.
No resolution.
No answer.
This thread documents the full timeline:
What was claimed.
What was tested.
What was found.
6
978
Verification Notes #24
"Verified" is quickly becoming a marketing label.
When a protocol claims verification, the real questions are simple:
- What was actually checked?
- Can someone else independently reproduce it?
Trust is a claim. Reproducibility is evidence.
3
50
Today I just want to say: thank you.
Thank you to the defenders of Ukraine for protecting our country.
We are living through an incredibly difficult time. And being able to work, speak, make plans, and simply wake up in the morning is not something we should take for granted.
Take care of yourselves. 🇺🇦
Glory to Ukraine.
Made with AI
4
59
Architecture remains a promise until an independent party can verify it.
@PsyProtocol still has a public 521k TPS benchmark, but the independent reproduction path remains unresolved.
Updated assessment:
github.com/Sergey007S/psy-pr…
“I didn’t allow that”
is becoming the sentence of the week.
Permission is a screen.
Architecture is a promise.
When money is in the loop:
What can the model never see?
Not your keys.
Not your authority.
You don’t fix that with a better settings page.
You fix it in the architecture.
4
57
Computation can happen anywhere.
The hard part is verifying it happened correctly.
That’s the real foundation of the cryptographic world computer.
The cryptographic world computer:
vitalik.eth.limo/general/202…
My attempt to express in somewhat concise terms the true meaning of basically everything planned to happen to Ethereum starting from the fork after Hegota. It's really not just a blockchain anymore. It's a hybrid architecture that combines together blockchains and modern cryptography, to enable much more powerful properties.
FOCIL, EIP-8288, Lean consensus, state management, formal verification, advanced mempool improvements (including privacy), and the longer-term specter of obfuscation all mentioned.
2
62
For decades, compliance meant collecting more data.
Now even regulators are discussing proving the required fact without collecting the underlying data.
That’s a much bigger shift than it sounds.
" A zero-knowledge proof can tell a counterparty 'Yes, this person meets your requirement' without that counterparty knowing your name, income, or address"
2026: SEC Commissioner @HesterPeirce suggests it.
2016: reviewer number 2 complains its unrealistic in a paper.
sec.gov/newsroom/speeches-st…
2
87
I'm pre-registered for the DAC mainnet launch. Mission card in hand — see you at the gate. @dac_chain interstellar.dachain.io/api/…
1
56
Independent verification starts where marketing ends.
I organized months of public findings regarding @PsyProtocol's public claims, benchmarks, and verification process into an open assessment repository:
• Technical assessment
• Public artifacts
• Timeline
• Industry engagement
• Supporting evidence
The objective is simple:
Document what is publicly reproducible and what remains an unverified assertion.
Independent reproducibility is not just a benchmark problem.
It may become a defining standard for evaluating public technical claims in Web3.
github.com/Sergey007S/psy-pr…
cc: @cmpeq @cyberqualia @StarkWareLtd @bcap
1
5
110
Verification Notes #23
Complexity is the enemy of verification.
Every hidden dependency is a place where trust quietly returns.
The goal isn't to remove complexity.
The goal is to make assumptions visible.
3
63
"Verify it" is easy to say.
The harder question is whether independent verification is actually possible.
If verification requires privileged access, trust hasn't been removed. It has only been relocated.
The future is verifiable:
Don’t trust the code. Verify it.
Don’t trust the identity. Verify it.
Don’t trust the agent. Verify it.
Trust less. Prove more.
2
71
The definition is becoming the critical security boundary.
AI may generate the implementation and even help prove correctness.
But the hardest question remains: are we verifying the right claim in the first place?
Invariants scale further than code.
2
57
Many investors speak about transparency, verification, and trust minimization.
What should happen when a portfolio company repeatedly promotes verification-focused messaging, while basic questions about independent reproducibility remain unanswered?
Should the burden of proof remain with the researcher, or with the project making the claim?
cc: @PsyProtocol @cmpeq @cyberqualia @StarkWareLtd @bcap
6
172
The first country to build a cryptographically relevant quantum computer may gain an advantage.
The first country to deploy post-quantum infrastructure at scale may gain the more durable advantage.
The race is not only about breaking cryptography.
It is also about replacing it.
⚡️ The race to quantum supremacy is accelerating.
The last time nations raced for a technology that could rewrite the balance of power, it ended in a space race and an arms race.
So is it realistic to expect them to cooperate on quantum?
Or does the first to build the machine keep it? 🤔
3
75
Every trust model has a root.
TEEs trust hardware.
Blockchains trust consensus.
Proof systems trust mathematics.
Security starts where assumptions stop being hidden.
4
66
Verification Notes #22
A timestamp proves bytes existed.
Explainability shows how a model reached a result.
Telemetry shows what a system says about itself.
None of them prove the result can be reproduced.
Visibility isn't verification.
Trust is a claim. Reproducibility is evidence.
4
72
Observation on Fast Sign:
"Sign a document. Anyone can verify it."
That's a powerful claim, but verification has layers.
Verifying that a hash exists on-chain is completely different from independently verifying the software, rules, and consensus that settled it.
Recording a fingerprint proves that bytes existed.
Independent verification proves how the system reached that result.
cc: @fastxyz @RRosmaninho1 @RosuGrigore
@cxh93 @polychain
4
126
An LLM can generate a candidate solution.
But research doesn't end when a model produces an answer.
It begins when someone independently verifies why it's true.
Generating outputs is getting cheaper. Verification is still the bottleneck.
Frontier LLMs can now solve interesting open problems. So how should academic conferences adapt? I get asked this often and I've given this a lot of thought for QIP 2027 specifically, as chair of the Steering Committee. My personal thoughts (not speaking for the SC): (1/4)
3
96
The interesting part isn't ZK itself.
It's that some of the world's best mathematicians seem to be converging on the same idea:
trust is not a security model.
notable that zk-proofs are the primary, and really the only, example given in this piece about mathematicians starting a new AI safety org. nytimes.com/2026/09/08/scien…
2
67
The marketing keeps moving. The verification gap doesn't.
@PsyProtocol is publicly promoting:
• 521k TPS as a verifiable benchmark
• private wallets and batched execution
• agentic wallets with permissions
• “trustless” infrastructure
• agent rules that are “verifiable on-chain” and “auditable by anyone”
The list of claims keeps growing.
The open question is still the same:
Can an independent researcher reproduce and verify these claims end-to-end from public artifacts?
Assessment: 33/100
Performance transparency: 0/100
Assessment:
Crypto Project Verification Analyzer
crypto-project-verification-…
Research:
Psy Protocol Assessment
github.com/Sergey007S/psy-pr…
cc: @cmpeq @cyberqualia @StarkWareLtd @bcap
1
5
94
Cards don't answer verification questions.
Merchant integrations don't answer verification questions.
"Verifiable finality" doesn't answer verification questions.
Since April, discussions around Verifier Quorum = 0 led to public and private conversations — but not to a publicly reproducible explanation of how an external party can verify an accepted claim.
Meanwhile, Fast is expanding the financial surface: agent payments, merchant checkout and cards.
More financial capability means a higher verification burden, not a lower one.
If the verification model changed, where is the evidence?
If it didn't change, what exactly is being verified?
Momentum scales adoption.
Independent verification scales trust.
cc: @fastxyz @RRosmaninho1 @RosuGrigore @_CHEN_Xiaohong_ @polychain
1
5
170
Grigore Rosu's team just published an impressive campaign: 158 codebases reviewed, K-to-Lean exports, and machine-checked proofs for SWE-bench.
The academic rigor is real.
While reviewing the newly released material, I noticed that both fast-wallet (production) and fast-wallet-design remain private repositories. The only public repository related to the audit contains UI review assets and screenshots.
As Rosu noted: "everything scientifically true is backed by a proof."
The open question remains: what public artifacts currently allow an external party to independently understand or verify the behavior of Fast Wallet in production?
cc: @fastxyz @RRosmaninho1 @RosuGrigore
@cxh93 @polychain
2
56