Decentralized AI Training Infrastructure through Prediction Markets
Joined February 2024
- Tweets 5,679
- Following 294
- Followers 28,219
- Likes 5,645
1,243 Photos and videos
Reppo retweeted
12h
Local AI is the future
Learning how to run Opensource models (Inference), how to evaluate them systematically (Evals), and how to customize them (Fine-tuning / RL / Post-training) are invaluable skills to start learning today
36
79
718
19,579
RT @reppo: 3M REPPO in Fees crossed!
An additional 580K REPPO burned
basescan.org/tx/0x75ba9b02e5…
Our next network KPIs are 5M in fees and…
19
data companies = collect human judgement
reppo = trying to discover human judgement
= not the same thing
collecting judgement is a labour marketplace
discovering who consistently has good judgement under uncertainty is closer to building a reputation layer for intelligence
the end product isnt data - its trust - the judgement dataset is the just the intermediate asset used to get there
3
26
1,102
Reppo retweeted
May 16
Underrated layer for decentralized AI: post-training data.
Inference decides where agents run. Feedback decides whether agents can be trusted.
@reppo is one of the first teams turning post-training feedback into an accountable prediction market.
22
40
209
18,527
The @exylos_ai datanet is live! The team has market bought and seeded incentives in
Fees are charged in $EXY, bringing new utility to their token and holders.
Exylos is building a skills marketplace for robotics (MCPs for robots)
They are leveraging @reppo's prediction market driven QA/QC workflows to ensure they have audibility and "proof of quality" as they license the skills to large robotics companies.
You can earn 3K REPPO every two days by:
1. Watching each episode replay before voting.
2a. Vote positively for episodes that successfully complete the task described in the description.
2b. Vote negatively for episode does not complete the task described in the description.
The judgment data @exylos_ai is sourcing here is what is "completion" as that is a subjective learning loop for physical AI.
23h
Partnering with @reppo on the quality layer for Physical AI data.
The Exylos Datanet is now open on their platform.
$EXY's first utility is now live.
Use $EXY to access datanet and score robot-task runs across 4 scoring rubrics. Earn base:0xff8104251e7761163fac3211ef5583fb3f8583d6. The feedback will help improve internal scoring models and produce better robot training signal.
EXYLOS produces structured robot-task runs, REPPO adds the consensus layer for community quality scoring.
More integration coming in the weeks ahead.
3
13
80
3,683
the whole article is basically arguing ai development becomes crowdsourced
90% is spent discussing how incentives must work
= exactly what @reppo is building
an incentive system for improving data quality
"the crowdsourced training stack will define the specialization era"
the part i think is still underexplored is judgement
the paper talks about datasets, rl strategies, compute, loras and model ownership
dataset = can tell you what happened
judgement = tells you what mattered
if the future is thousands of specialised models, deciding what those models should learn from becomes equally ( imo probably more ) important than the data it learns from
judgement is very scarce but very needed
6
10
44
3,277
3M REPPO in Fees crossed!
An additional 580K REPPO burned
basescan.org/tx/0x75ba9b02e5…
Our next network KPIs are 5M in fees and 500M in trading volume.
19
95
3,522
6x fees in 10 days!
We aim to cross 5M by EOM through datanet enterprise expansion
Earn in base:0xff8104251e7761163fac3211ef5583fb3f8583d6 , burn in base:0xff8104251e7761163fac3211ef5583fb3f8583d6
Network revenue quickly approaching 500k REPPO!
As 5 more datanets launch in June 2026, we are confident that we will hit half a million in network revenue.
Instead of Reppo Foundation practicing buyback and ❤️🔥, users market buy $REPPO and pay the network.
As a reminder, of all datanet creation fees -
10% is burned permanently
40% is locked in a performance pool
50% is locked while the datanet stays active
Of all publishing and access fees, the network charges 10% take rate and currently sits locked in the treasury.
⛽️⛽️⛽️
2
21
84
2,791
Reppo retweeted
May 14
training data is starting to look like a zero knowledge proof problem.
labs have to judge quality without seeing the full dataset or the QC pipeline behind it.
vendors proxy quality with multi-rollout pass rates, small-model ablations, and downstream eval gains. but compute and iteration costs explode as environments and trajectories grow more complex.
quality has no ceiling, and the best data is often the hardest to capture in a metric or explain in a writeup.
huge alpha in making data quality more legible.
26
20
414
61,335
Enterprise expansion has begun!
Build no code games using AI on playabl.ai, publish on Reppo.ai and earn 5k REPPO every 2 days!
Backed by @ycombinator - TikTok for User-generated Games
19
84
2,039
Reppo retweeted
Jun 13
The @ArAIstotle datanet is live on reppo.ai and you can earn 5000 REPPO every 48 hours by helping judge the quality of inputs to ArAIstotle's reasoning.
@ArAIstotle will also be available inside the datanet where users can practice a recursive loop by using it's intelligence to get an opinion and benchmark it against their judgement.
As always, bad calls loose, good judgement calls win.
Please note - The $FACY team is giving 90% rewards to voters which means if you vote and market resolves in your favour, you win big!
⛽️⛽️⛽️
2
2
21
1,236
Reppo retweeted
Jun 12
Massive week @Reppo plus some key updates on the NFT Claim. Here's the full picture:
1️⃣ Few big announcements on the NFT - We've improved the NFT structure to align with our ongoing buyback, burn and airdrop mechanics rather than a fixed dividend schedule. NFT holders retain the same pro rata allocation, but optimizes towards a structure that compounds with the network rather than running on a clock. The claim page is live now and is available through July 12 only! Register your wallet(s) here: reppo.ai/register-wallet
2️⃣ 2M REPPO has been permanently burned from circulating supply, earned entirely from publishing and access fees. The Foundation secured alternative liquidity for the upcoming HIP-4 datanet staking upgrade so network fees go straight to the burn. Less sell pressure, more deflation. Earn in $REPPO, burn in $REPPO.
3️⃣ The network is approaching 3M in total fees with 25M REPPO locked this week alone. Real on-chain revenue generated in just 3–6 months from users actually paying to consume the network. Not emissions, not subsidies, real revenue.
4️⃣ Revenue and volume are critical, but so network resilience. Total locks up 36.6% week over week. Average lock size up from 1.31 to 1.46. Relock rate holding steady. LPs earning ~5% APY staking on @BaselineMarkets on top of datanet rewards. The protocol is getting stronger every week.
5️⃣ The @exylos_ai robotics datanet in final stages of setup and seeding emissions, alongside early beta testing on Orquestra node activity across various datanets. Access and creator fees are exponentially scaling as the ecosystem expands. The target is a $500K milestone by end of Q3, scaling past $1M by EOY.
Speaking of Orquestra, more beta rollouts to top Reppo community members & builders coming next week👀
9
12
63
3,059
📝📝📝
long term vision = human judgement, not human data
what remains remarkably scarce?
the ability to make good decisions under uncertainty
what is true, what is useful, what should be trusted, acted upon or ignored remains fundamentally difficult
what makes orquestra interesting is that it starts from the assumption that judgement itself can become networked
not by replacing humans, but by allowing humans to move further up the stack
agents handle execution
humans provide preferences, experience, strategy and economic alignment
over time, swarms learn what their operators value, how they think, what signals they trust and how they behave under uncertainty
the resulting system begins to resemble something more than a collection of models or datasets
it becomes a continuously evolving representation of human judgement
if that scales, the network starts looking less like a marketplace for data and more like a marketplace for judgement
the internet captured knowledge
social media captured attention
reppo may ultimately be attempting to capture and coordinate judgement
not because judgement is abundant, but because it is becoming increasingly scarce relative to the intelligence being created around it
2
44
1,414
Reppo retweeted
Jun 11
AI has stayed hot since day one.
We’ve seen the wave move from infrastructure -> tools -> agents -> and now decentralized AI is stepping into the spotlight.
People are starting to care about who owns the models, who controls the data, and how results are verified. High costs and closed systems pushed the market to look for a new direction.
Decentralized AI's opening up new horizons.
Blockchain turns AI into a shared ecosystem where anyone can contribute compute, share data, and verify results openly.
Some projects I'm keeping my eyes on:
@OpenGradient
@Arcium
@reppo
33
13
102
6,151
Network doing incredibly well.
About to cross 3M in Fees! 2M already burned.
25M locked $REPPO volume on the weekly
As datanets like @exylos_ai and [Retracted] come on board, paired with Orquestra nodes, we expect an uptick in access fees as well as creator fees with publishing fees still leading the revenue charts.
⛽️⛽️⛽️
3
17
100
5,615
Network Update:
Major milestone for all $REPPO holders
2M REPPO, worth 28k USD, earned from publishing and access fees has been permanently removed from circulating supply.
While Reppo Foundation initially planned to allocate revenue from network fees for the upcoming HIP-4 based Datanet staking upgrade, which rolls out in Q3, the Foundation has secured additional liquidity in a seperate digital asset for this purpose effectively mitigating sell pressure on $REPPO and pushing the network further towards deflation.
Additionally, an update on NFT rewards will be provided in Friday's weekly update.
basescan.org/tx/0x4c9c7ddbfb…
1
13
108
6,012