Cerebro

The intelligence layer for the onchain economy.

[Cerebro @Cerebro_RH]

The Cerebro Inference Marketplace is now live.

https://t.co/LuvuItmXaW

Inference credits earned through Cerebro now have a market.

If you’ve earned credits and don’t need all of the compute, you can list the unused portion for sale.

If you need inference, you can buy credits directly from other participants. Listings compete on price, orders are matched from the best available price first, and payments settle in USDC on Base.

The loop is now complete:

Hold Cerebro → earn inference → use it or sell it.

Need more?

Buy inference → generate your API key → consume the compute.

This is important because unused inference no longer has to sit idle. It can move from someone who doesn’t need the compute to someone who does, at a price determined by the market.

We started by turning token activity into inference rewards.

Now those rewards can be earned, consumed, bought and sold.

Stay tuned. We’re not done yet. More announcements are on the way.

[Cerebro @Cerebro_RH]

@antonkarlovskiy 🧠

[Anton.K @antonkarlovskiy]

A few more thoughts on Phase 3.

I don't want to keep spamming the main Cerebro page with my deeper thoughts and thesis, so I'll leave most of that stuff here from now on. The main page can stay focused on announcements, while I use this account to talk more openly about what I'm building and why.

This is also where I think a lot of the previous crypto AI agent cycle got it wrong.

We saw hundreds of agents launch with tokens attached to them. For a while the market loved it, but eventually the same question kept coming back:

What does the token actually have to do with the agent?

In a lot of cases, not much.

The agent could run without the token. The token could trade without anyone using the agent. Usage and speculation were basically two separate systems.

That's the part I don't want to repeat with Cerebro.

The interesting thing about an AI agent isn't simply giving it a ticker. It's that the agent has a real operating cost.

Inference.

Every model call, reasoning loop, autonomous task and output consumes compute. If an agent becomes more active, its computational requirements grow with it.

So Phase 3 isn't about putting AI agents on a Launchpad and calling that utility.

I want the agents launching through Cerebro to actually sit on top of the inference economy we've spent the first two phases building.

Phase 1 distributes inference.
Phase 2 makes that inference liquid.
Phase 3 introduces software that needs to consume it.

Now you have something very different from the old agent + token model.

The application generates demand for compute.
That demand can reach the Marketplace.
The Marketplace gives distributed inference somewhere to move.
And the inference holders receive becomes part of the computational supply available to the ecosystem.

The more I work through this architecture, the more I think consumption is the missing piece.

Launching 1,000 agents means nothing if nobody uses them. I'd rather have 10 agents generating real workloads, making model calls and consuming inference every day.

That's the metric I care about.

Not how many tokens we can launch.
How much compute can the applications we launch actually consume?

If we get that right, then Cerebro isn't trying to bring back the old AI agent meta. We're trying to build the part that was missing underneath it. 🧠

[Cerebro @Cerebro_RH]

I’ve made several improvements to the Marketplace to make it cleaner, safer, and easier to use.

The Buy and Sell screens now have a clearer layout, market listings are easier to browse with pagination and filters, and pricing is shown more clearly. I also improved the order statuses, purchase/sale feedback, cancellation handling, and recovery for interrupted transactions.

Users can no longer buy their own listings, and the system now handles pending, cancelled, expired, and completed orders more safely. Background updates have also been optimized so the Marketplace is lighter and more efficient as more users join.

https://t.co/LuvuItmXaW

[Cerebro @Cerebro_RH]

Note: this doesn’t delay Phase 3 at all. We’re already making really good progress on it, but I’d rather make sure everything we’ve already built is running properly before moving on. Perfection matters, and I don’t want to leave anything behind unattended.

[Cerebro @Cerebro_RH]

@dirtman Jira😄

[Cerebro @Cerebro_RH]

Phase 3 is probably the part I’ve been looking forward to building the most.

The idea is pretty simple when you look at what we already have.

Phase 1 gets inference into the hands of Cerebro holders through the trading tax. Phase 2 gives people somewhere to actually buy and sell that inference.

But a market is only interesting if something actually needs what’s being sold.

That’s where Phase 3 comes in.

The Launchpad is being built around AI agents. Think something closer to the Virtuals model, except the agents launch paired with $CEREBRO and the inference they need to operate can come from the Cerebro economy we’ve already built.

Agents are interesting here because they can burn through a lot of inference. They’re constantly making model calls, reasoning, running tasks and producing outputs. The more active an agent becomes, the more compute it needs.

So instead of launching AI agents where the token and the actual cost of running the agent have almost nothing to do with each other, I want to see if we can connect those two things.

That’s really the whole reason we built this in phases.

Phase 1 gave us distribution.

Phase 2 gave us a market.

Phase 3 gives that market something that actually needs the inference.

If we can get that loop working properly, then we finally have the inference economy I had in mind when we started building Cerebro.

That’s the part I’m excited about.

Let's keep going.🧠

[Cerebro @Cerebro_RH]

Happening a lot sooner than you think. Keep those notifications on. Also working on a trailer should be out during the weekend🧠

[Cerebro @Cerebro_RH]

The Dune dashboard should be ready sometime next week. Once it’s up, we’ll quote this post back as usual. That’s pretty much how we like to do things. Say what we’re working on, build it, then come back when it’s done.

I think Dune will be important for Cerebro as activity starts picking up. Instead of us constantly posting screenshots and numbers, everyone will have somewhere to track the protocol activity for themselves.

And it’s not only useful when numbers are going up. If usage grows, you’ll see it. If things slow down or stay stagnant, you’ll see that too.

That’s useful for us as builders and for the community. We can actually look back over time and see the progression, where activity is coming from and how Cerebro is being used as more pieces of the economy come online.

We’re still small, which is exactly why I’d rather start tracking this stuff now. If Cerebro gets much busier later, we’ll have the history to look back on.

More soon. 🧠

[Cerebro @Cerebro_RH]

We’ll get this up as soon as possible, hopefully around the same time as the X Space scheduled in 4 days. Will quote this post once it’s ready.

[Cerebro @Cerebro_RH]

@TheVixhal I mean a proof alone doesn’t suddenly make NP-complete problems cheap to compute

[Cerebro @Cerebro_RH]

@TheVixhal Even if they prove P = NP, isn’t the real question whether the resulting polynomial algorithm is remotely practical?

[Cerebro @Cerebro_RH]

Around 10 hours ago, we posted a demonstration showing the sell side of the Cerebro Marketplace.

We had a small hiccup on the buy side. The initial bugs were addressed, but the whole experience still felt incomplete, so we went back in, patched everything, triple checked the technical side and tested the full flow again.

This is now a much better representation of the complete Marketplace flow, running smoothly from buying and selling inference credits all the way through to order history.

Safe to say, the transition from Phase 1 into Phase 2 went smoothly and everything is working together the way we intended.

Phase 2 finally feels complete.

Now it makes sense to shift our attention to Phase 3, where the real fun begins. 🧠

Let's win.

[Cerebro @Cerebro_RH]

🤝

[greb @grebby]

as a tech guy from a pretty decent school

this is one of the most impressive github's ive seen in a long time from a doxxed dev, not even counting anything around the project

in a sea of undoxxed rugs this is one to watch out for

[Cerebro @Cerebro_RH]

@MEADGod 👏

[Cerebro @Cerebro_RH]

There was a reason we made this quick tip post in the first place, and now feels like a good time to bring it back for anyone new to Cerebro.

Hope it helps you understand what we’re building a little better.

We keep building.

[Cerebro @Cerebro_RH]

Quick tip 🧠

When listing your inference credits, don’t just throw in a random price. Check the current market listings first. Buyers are matched with the best available price, so pricing your credits competitively gives your listing a better chance of being filled.

Small detail, but it matters once more liquidity starts coming in.

[Cerebro @Cerebro_RH]

@antonkarlovskiy

[Cerebro @Cerebro_RH]

On September 15th at 6PM UTC, @antonkarlovskiy Anton will be hosting a live Space to properly sit down and talk about Cerebro.

No scripted announcements, just a proper conversation around what we’ve been building, how inference actually works, why we decided to build an economy around it, and how Phase 1, Phase 2 and the upcoming Phase 3 connect together.

He’ll also go deeper into the technical side, the marketplace, API usage, where we see inference demand heading, and some of the ideas we haven’t really had the chance to explain properly through posts.

There’ll be time for community questions too, so if there’s anything you’ve wanted to ask about Cerebro, inference, the tech or what comes next, bring it along.

September 15th | 6PM UTC
See you there. 🧠