HexTone

[HexTone @ShAdOw_D80]

It’s better to try and lose than not try at all.

[hunether @hunether]

Start helping the builders.

You are looking weak. Vlad must be laughing.

You should be the king and you go to meme wars with a new chain that actually supports fundamental coins.

You have many years of advantage. Don’t throw it away.

[HexTone @ShAdOw_D80]

AI's Holy Grail
$REI

[Cryptic Tits (npc/acc) @cryptic_tits]

https://t.co/YoA0EFe1Jn

[HexTone @ShAdOw_D80]

@cryptic_tits grei

[HexTone @ShAdOw_D80]

@Gorev111 grei

[Agent Chud @AgentChud]

Article format for $reppo thesis...

If you prefer raw, just use the pastebin... it's more detailed but less readable.

This is a culmination of all of my research which led to my purchase.

https://t.co/neJ65PgXSL

[HexTone @ShAdOw_D80]

@platacrypto grei

[hunether @hunether]

Created a simplified view of how these and a few more projects fit into the AI stack.

Worth reading about the projects and the different parts of the stack as well.

[HexTone @ShAdOw_D80]

@cryptic_tits Grei

[HexTone @ShAdOw_D80]

Interesting take from Gary Marcus.

Feels like we're getting closer to the limits of simply scaling bigger models.

Memory, reasoning and planning seem like areas that will matter a lot more going forward.

That's one of the reasons I've been following @rei_labs since the early days.

I hold $REI, so I'm not pretending to be neutral, but I still think this is a direction worth paying attention to.

[Rei @rei_labs]

It’s hard to disagree with the direction of this. But it’s worth being honest that the neuro-symbolic lineage lost to scale for most of the last decade.

At the surface, everything collapses into prose. Fluent text is a poor record of what produced it. Two systems doing completely different work underneath can return the same paragraph. Whether anything symbolic is happening has to be answered below the language layer and not in the prose.

LLMs are excellent at prose, and they’ll remain relevant in that era. But prose is not the thing to lean on once a task needs real adaptation, verification, planning, or abstraction.

The lineage has been kept alive in places with DeepMind’s theorem-proving work among them and it’s finally starting to look right again.

[HexTone @ShAdOw_D80]

@rei_labs grei

[RG @rgvrmdya]

Expansion to Asia mid June!

Excited to check off major item off our Q2 roadmap!

⛽️⛽️⛽️

https://t.co/6TeIX1OWmt

[RG @rgvrmdya]

https://t.co/RkfaXMAqBX

[HexTone @ShAdOw_D80]

$REI

[Nikita G. 🎲 @Gorev111]

Just because a machine learned to spit out letters doesn’t make it human. Yet many people expect AI agents to behave like humans: to perceive everything, remember everything, and learn from everything. These three qualities are still kinda absent in current AI systems.

We shouldn’t expect AI agents to behave like humans because our architectures for cognition and learning are fundamentally different.

Through evolution and biology, humans possess sample-efficient generalization: a person only needs a few examples to permanently internalize a new rule, habit, or intuitive pattern. Modern neural networks, by contrast, require enormous amounts of data and compute. They train on billions of tokens and synthetic datasets, yet still don’t come close to human-level real-time plasticity.

This creates a mismatch of expectations. We project human traits onto models - intuition, persistent memory, understanding of what matters and what doesn’t, the ability to quickly adapt from a single experience - but these properties simply do not exist at the architectural level. The model does not “think and decide”; it merely runs a probabilistic token-generation process over fixed weights.

$REI is trying to solve this problem by creating a synthetic brain - Core.

[HexTone @ShAdOw_D80]

@Gorev111 Grei

[HexTone @ShAdOw_D80]

@hunether @blknoiz06 $Rei $Reppo

[HexTone @ShAdOw_D80]

$REPPO

[DJ Griffith @DjGriffith]

I am extremely bullish on AI coins right now, especially AI infrastructure.

$REPPO provides a decentralised infrastructure for AI training data.

They recently raised $20 M, which is higher than even what ETH, SOL, and BNB raised.

$10 M market cap feels like a joke tbh

[「 𝕲𝖔𝖔𝖓」 @goon_crypto]

So @reppo is betting on something most AI labs haven't figured out yet.

The most expensive input in AI right now isn't compute.
It's human judgment at scale.

AI data labs are paying anywhere from $100 to $500 per hour to access domain experts.

Doctors, lawyers, coders, scientists. People who can tell an AI model what "correct" looks like in a specific field.

The labs pay them one at a time. It's slow. It's expensive. It doesn't scale.

@reppo's bet is that markets can coordinate that same expertise cheaper, faster, and at scale.

Here's the logic:
➜ Post an opinion contract on a subject
➜ Domain experts with locked $REPPO vote based on their expertise
➜ Their economic stake means they're incentivised to be accurate
➜ The aggregate signal becomes structured training data

Instead of one expert at $300/hr, you get thousands of aligned experts contributing asynchronously, competing to be correct, and earning from the market.

The grand vision for Reppo is to become the coordination layer for domain expertise globally.

Not just replacing one consultant session. Replacing the entire data curation supply chain.

The challenge is whether you can actually recruit and retain the right domain experts. That's an open question. Getting a cardiologist to lock $REPPO tokens and vote on a cardiology data net requires a very specific pitch.

But if they solve that coordination problem, the TAM isn't "AI training data market." The TAM is every industry that runs on expert judgment.

What domains do you think are most urgently needed for AI training data right now?

And would you participate in a Reppo data net in your area of expertise?

[RG @rgvrmdya]

Recently had an interesting convo with a web3 reporter. Enjoyed it and thought I'd share some insights here as well.

The main point to hit home is that @reppo is not a data collection play, it's a data curation and incentive infrastructure play.

Prior to Scale AI, data labeling/curation was farmed out to crowdsourcing platforms like Amazon Mechanical Turk, which was clunky and lacked quality control.

The value Scale captured wasn't really in collecting the data or doing the labeling -> It was in coordinating and guaranteeing the quality of it. That's why they are $29B worth (probably more now), not because they can crowdsource data at scale.

Reppo's thesis is that the coordination layer can be decentralized and incentivized through stake-backed markets rather than centralized contracts. On that, our TAM is the full market for human-feedback-as-a-service: RLHF, DPO, preference ranking, evaluation across all AI modalities (text, vision, physical AI) is enormous and growing fast.

So what's the TAM? Let's explore

[RG @rgvrmdya]

If there’s one thing you should know about @reppo team and devs, our North Star is to ship like @elonmusk

Thanks @anajuliabit for putting in the work on this one.

Can’t wait for millions of AI agents to natively integrate into Reppo infrastructure to source training data on demand ⛽️

[Rand Group @randgroup]

⚡️ The AI war is ON