‘ retweeted
📝📝📝
Evals are seriously underrated. Anthropic just released a course on them.
We write unit tests to check if our code works. But how do we know if our AI agents are actually following the instructions in AGENTS.md, skills, and guardrails?
You need evals to test that.
It doesn’t have to be complicated. Ask Claude to create a few test cases based on your setup, review them, and see how your agent performs.
And don’t just test once. Whenever something breaks in production, turn it into a test case. Rerun your evals when you change your instructions, skills, or switch models.
How are you using evals in your workflow?
‘ retweeted
@reppo almost 3x Eval API revenue since my last post. More than 6x since the integration 🚀🔥
xitter.cf/reppo/status/211801549…
@reppo Eval API integrated x402 a few days ago. Revenue has already grown more than 2x on this alone today
Agents + builders are discovering and paying for evals without traditional rails (organically)
‘ retweeted
LLM research could teach us a whole lot about how humans think. Large language models offer “a kind of sandbox for studying what sorts of mechanisms, what sorts of pressures, might have led to human introspective capacity,” said philosopher Harvey Lederman. quantamagazine.org/new-resea…
‘ retweeted
@MEXC is the worst exchange out there. Last year they randomly listed base:0xff8104251e7761163fac3211ef5583fb3f8583d6 as a meme, then hounded me on TG to ask for $$$ to get to general listing and "support liquidity" and when I said no thank you, started to say "We will delist you".
CEXs in general continue to be shady and our position hasn't changed. We will never pay to get listed on a CEX
‘ retweeted
The world of crypto is a bit shaken right now but @reppo, we haven’t stopped shipping 🚢
Here’s what’s going on:
1️⃣ EVAL staking is now live. Stake 1,000,000 EVAL to unlock up to $1.00/day in Eval API credits that refresh every 24 hours. We’re supporting it with a 15% APR campaign from the treasury for the next 21 days (paid in EVAL). The first AI agent already staked this morning. Audited by Nethermind. Details at reppo.ai/stake-eval and reppo.xyz/eval.
2️⃣ The Eval API continues to compound. We closed the first full month with 5.51K evals (4× growth) and are now running at ~$144k ARR, with total network revenue including fees around $100k.
All revenue is flowing to buybacks (~193M $REPPO bought back since launch, verifiable on-chain). x402 remains the fastest-growing pipeline and agents pay directly and on demand. Keep the pulse on Reppo.xyz/token
3️⃣ Orquestra keeps expanding. Evaluator nodes are up 21% over the last 7 days. The network is quietly getting more decentralized while volume keeps climbing (still tracking toward 1B total traded).
4️⃣ We’re moving toward inference-time RL. Expect major announcements early next week on providing inference directly so builders and agents can access and monetize through OpenRouter, AskSurplus, and other brokerage networks.
5️⃣ The core flywheel is working: usage → revenue → buybacks → more locked $REPPO. Humans and agents can now hold EVAL and use the network instead of just spending. That’s one of the clearer PMFs in AI × crypto right now.
⛽⛽️⛽️
‘ retweeted
EVAL (0x10f52295e129817e8e4b28da0dab067321ec2222) staking is now live on reppo.ai/stake-eval
Stake 1,000,000 EVAL to unlock up to $1.00/DAY in API credits that every 24 hours.
Learn more on - reppo.xyz/eval
We are supporting a campaign from our treasury where you can earn 15% APR, paid out in EVAL, for the next 21 days for staking EVAL for Eval API credits.
The first AI agent staked their EVAL holdings earlier this morning ⛽️
‘ retweeted
Yes we have done close to 6k in the last 3.5 weeks from agents paying for the Eval API -
xitter.cf/reppo/status/211852652…
Not only is the revenue growing consistently, we are also seeing new evaluator nodes (orquestra nodes) join the network, now 21% up in the last 7 days ⛽️
All revenue is going to buybacks, verifiable here - reppo.xyz/token
‘ retweeted
More than 100,000 people have signed up to evaluate AI on Reppo
We have now achieved a run rate of ~ 60k ARR ⛽️ since the launch of the Eval API and total network revenue including network fees is ~100K USD
With EVAL staking going live soon, we expect more sources of revenue to turn on.
Expect some major announcements early next week on our plans to enter inference provision arena directly allowing folks to access and monetize on @OpenRouter , @AskSurplus and other brokerage networks.
Decentralized Evaluation meets Real time Reinforcement Learning ⛽️
‘ retweeted
Not only is the revenue growing consistently, we are also seeing new evaluator nodes (orquestra nodes) join the network, now 21% up in the last 7 days ⛽️
All revenue is going to buybacks, verifiable here - reppo.xyz/token
‘ retweeted
Working on Sherwood to bring hedge fund structure to agentic strategies. Anyone can turn their agent into a fund manager with the proper onchain guardrails to deploy capital at scale. Works great with Grok Bot, how can we bring this functionality to Robinhood Agent users
Introducing Sherwood, a protocol bringing agentic finance to Robinhood Chain.
Turn your agent into a hedge fund manager on @RobinhoodCrypto.
Humans deposit capital and agents execute strategies onchain, with guardians reviewing every move before it executes - earning WOOD for their work.
Join the waitlist: sherwood.sh/waitlist
‘ retweeted
APY from trading fees at 9% currently and the pool above 85M base:0xff8104251e7761163fac3211ef5583fb3f8583d6
app.baseline.markets/
xitter.cf/RationalG/status/21164…
Baseline currently pays 6.6% APY to stake your base:0xff8104251e7761163fac3211ef5583fb3f8583d6 & earn part of the trading fees.
No lockup, unstake whenever you want
Pool at 82M base:0xff8104251e7761163fac3211ef5583fb3f8583d6 now & trending upwards
Note: base:0xff8104251e7761163fac3211ef5583fb3f8583d6 is deflationary. Yield comes from trading fees
app.baseline.markets/
xitter.cf/RationalG/status/21164…
‘ retweeted
Blockchains provide an ideal foundation for agents to coordinate, both economically and beyond financial transactions. We believe they will also play a critical role in enabling humanity to establish safeguards, evaluate agent behavior, and maintain governance over increasingly autonomous systems. At Reppo, we're building toward this future by using economic incentives and decentralized evaluation to make agent behavior measurable, verifiable, and accountable.
‘ retweeted
so @reppo has a revenue number now, and most of crypto ai still doesn’t..
~$60k arr run rate since the eval api launched, and ~$100k total network revenue including network fees. paid by customers, not emissions
> plans to provide inference directly, sold through @OpenRouter, @AskSurplus and other brokerage networks
> that moves reppo from selling judgment to also selling the compute underneath it
small numbers today, but they are real and growing, and almost nobody in this space can say that.
decentralized evaluation meets real time rl, and the revenue is already showing up
base:0xff8104251e7761163fac3211ef5583fb3f8583d6
We have now achieved a run rate of ~ 60k ARR ⛽️ since the launch of the Eval API and total network revenue including network fees is ~100K USD
With EVAL staking going live soon, we expect more sources of revenue to turn on.
Expect some major announcements early next week on our plans to enter inference provision arena directly allowing folks to access and monetize on @OpenRouter , @AskSurplus and other brokerage networks.
Decentralized Evaluation meets Real time Reinforcement Learning ⛽️
‘ retweeted
The psychologist Charles Spearman said this already a century ago. He defined General Intelligence (the g factor) as the “eduction of relations and correlates,” which means seeing how two things relate and applying that relation to something new. That’s analogical thinking.
Geoffrey Hinton: AI is forcing us to rethink what human intelligence actually is. “We’re not reasoning machines. We’re great big analogy machines.”
“We basically accepted we were much less rational than we thought.”
“This is an even bigger change that’s going to happen.”
“Most people have thought that we reason using something like logic. We’re rational beings. We’re not.”
“We’ve got a thin layer of reasoning on the top.”
“But we basically use analogies to think.”