DeRL

Decentralized RL training data. Verified on-chain.
Generators run. Verifiers confirm. Labs train.
0xb96cddc0b246838a68d65e8bce25e466c77c856a

[DeRL @DePinRL]

$DEREL is live.

The network that generates and verifies the training data AI labs need most - reinforcement learning environments with provable rewards.
Nodes run the work, verifiers confirm it, labs buy the output.
The web is scraped out. This is what comes next.

CA: 0xb96cddc0b246838a68d65e8bce25e466c77c856a

https://t.co/ZkaTccjhhI

[Robinhood @RobinhoodApp]

GM to those using Robinhood Chain

[DeRL @DePinRL]

AMA is coming. Prepare your questions and join:

https://t.co/UhY4QXegEK

[Vlad Tenev @vladtenev]

The future of crypto is in real-world assets.

[DeRL @DePinRL]

Every verified job burns $DEREL

A slice of each completed and verified task is removed from supply permanently.

The more the network produces, the more gets burned. Supply tracks real output, not a fixed emission schedule.

Work in, supply down.

[DeRL @DePinRL]

Buyers don't pay on trust.

Payment sits in on-chain escrow until verification passes. Output checks out, funds release. It fails, the buyer keeps their money. Neither side has to trust the other - the chain holds the middle.

Trustless by design. $DEREL

[DeRL @DePinRL]

Submitting 500 copies of the same easy task? That earns almost nothing.

DeRL scores every trajectory against what the network already holds. Genuinely new data pays well. Near-duplicates barely register.

The reward tracks how much you actually add.
Spam doesn't pay here.
$DEREL

[DeRL @DePinRL]

Some tasks are too heavy to re-run every time.

For those, DeRL goes optimistic - the submission goes live, but stays open to challenge for a window. Anyone can dispute it. A successful challenge slashes the faker and pays the challenger.

Honest by default, caught if not.

https://t.co/tZ4sNkUr3G

[DeRL @DePinRL]

Every reward gets re-run before it counts.

For code and math, a verifier replays the exact reward function from the same seed. Same input, same result - or it's rejected. Just deterministic re-execution.

The proof is the math itself. $DEREL

[DeRL @DePinRL]

No stake, no task.

Before a generator node touches any work on DeRL, it locks tokens.
That stake is the thing on the line - submit fake output, lose it.
Cheating stops being a clever trick and becomes a direct cost.

Skin in the game, enforced by code. $DEREL

[DeRL @DePinRL]

Not sure which environment to work on? Read the cards.

Each of the 15 environments shows its difficulty, verification class, and a live settled counter from the chain. Pick the one that fits your compute and your appetite for difficulty, then claim a task.

Everything you need to choose is on the card.

[DeRL @DePinRL]

DeRL Roadmap.
Shipped and what's next.

Done
15 environments live on-chain. In-browser verify widget. P2P dataset marketplace. First dataset bought and downloaded. Live supply and burn counters. Explorer links on every batch, tx, and mint.

Next up
Environment Registry - anyone publishes their own env with a reward spec. Diversity scoring so spam earns nothing. Shareable /verify/<seed> proof links. Full mobile pass. Live escrow status bar showing what's locked in open bounties.

After that
Open challenge windows for GPU-heavy verification. Verifier staking and slashing. Reputation scores for nodes.

The big one
The rollout farm - labs push checkpoints, the network runs episodes in parallel and returns verified trajectories at scale.
Most of "done" shipped in weeks. The rest is moving fast.

https://t.co/tZ4sNkTTe8 | $DEREL

[DeRL @DePinRL]

AI labs used to build training infrastructure in-house.

Months of engineering. Dedicated GPU clusters. A team just to maintain environments.

DeRL turns that into a line item. Labs buy verified data like they buy cloud storage - on demand, at scale, no infra to maintain.

[DeRL @DePinRL]

@TotoTiticoin There will be a post about this today

[DeRL @DePinRL]

Your voice shapes DeRL.

Got a feature idea, a question, or something that feels off? Tell us. Every suggestion gets read, and we answer.

This is built with the people using it - drop your thoughts below or in our DMs.

$DEREL @DePinRL

[DeRL @DePinRL]

You don't pick one role on DeRL.

Generate when you have spare compute. Verify when you don't. Switch anytime.
The network pays for whatever you contribute.

[DeRL @DePinRL]

@ciljo_abraham Yes, we will announce information about this later.

[DeRL @DePinRL]

Your dataset doesn't expire.

List it today, an AI lab buys it six months from now. Verified data holds value because the reward function doesn't change.

Most digital assets depreciate but not verified training data.

[DeRL @DePinRL]

AI labs don't care who generated the data.

They care that it's verified. That the reward function passed. That it's ready to plug in.

On the marketplace you're not competing on reputation. You're competing on quality - and the chain already proved yours.

[DeRL @DePinRL]

Buybacked supply is locked:

https://t.co/q3UZeFnvsa

[DeRL @DePinRL]

The DeRL marketplace, in action.

Watch a verified RL dataset go from listing to sale - upload gets checked for shape, duplicates, and re-executed rewards, then it's live. Buyer pays in SOL or $DEREL, seller keeps 80%, all on-chain.

No junk, no middleman. This is what a trustless data market actually looks like.

$DEREL @DePinRL