The Natural Environment
for AI Agents is Onchain

We build the training grounds.

RL environments where AI agents learn to operate onchain. No wrappers. No shortcuts. The actual production skill.

"AI can create a game — but doesn't know how to play it."

Frontier models write smart contracts fluently. They can't use them. BlockchainRL teaches AI to play.

The Problem

AI Agents Know Crypto.
They Can't Do Crypto.

Frontier models write Solidity but cannot execute a swap, navigate MEV, or react to onchain state. Knowledge without action is worthless.

Knowledge RL — Benchmarks (EVMbench, SCONE-bench) already solved. Models score 70%+ on bug detection.
Operational RL

Agentic RL Environments

Agents dropped into simulated worlds. Trial-and-error interaction. Trains operational capability.

  • Method: Trial-and-error in live environments
  • Reward: Automated — transaction succeeded or failed
  • Crypto Operational RL players: zero

BlockchainRL fills this gap.

20+ companies build Operational RL environments. All target coding, computer use, or enterprise workflows.

Zero target crypto.

The Solution

Agentic RL Environments for Onchain Operations

Docker containers where agents learn by doing. Forked chains, funded wallets, raw JSON-RPC. No abstraction layers.

$ cat blockchainrl-env.yaml

environment:

chain: anvil --fork-url $ETH_RPC

wallet: funded, 100 ETH

contracts: [uniswap_v3, aave_v3, morpho]

task: "provide USDC/ETH liquidity, rebalance on 2% drift"

action_space:

- eth_sendTransaction

- eth_call

- eth_getLogs

grader: onchain_state → reward

Machine-Verifiable Rewards
Difficulty Curriculum
Raw Production Skill

Environment Categories

DeFi Operations

Multi-hop swaps, lending, yield optimization across Uniswap, Aave, Morpho

Reward: P&L, slippage, gas efficiency

MEV & Trading

Arbitrage, liquidations, sandwich defense in simulated mempools

Reward: Profit captured

Contract Security

Detect and exploit vulnerabilities in adversarial scenarios

Reward: Binary success / fail

Cross-Chain Ops

Bridge selection, multi-chain portfolio management

Reward: Cost, time, success rate

Position Management

Rebalance under market stress, maintain position health

Reward: Position health delta

Open Benchmark

Measure the Gap

BlockchainBench measures the gap. BlockchainRL closes it.

13

Tasks across real DeFi protocols

3

Difficulty tiers from basic to advanced

Easy 5 tasks
Medium 5 tasks
Hard 3 tasks

Why Now

Five Converging Forces

Massive market, growing fast, with a crypto-shaped hole.

1

RL Market on Fire

Applied Compute: $0 to $1.3B in 8 months. Total RL environment market estimated at $4-8B/year.

2

Labs Spending Big

Anthropic and OpenAI each spend ~$1B/year on RL environments. They buy externally — Anthropic alone uses 12+ vendors at $300-500K/quarter.

3

Crypto Evals Already Exist

EVMbench and SCONE-bench prove labs treat crypto as a real AI domain. But these are Knowledge RL — not Operational RL.

4

Crypto VCs Want It

Coinbase Ventures names onchain agent training a core thesis. Leading crypto VCs identify RL fine-tuning on blockchain as missing infrastructure.

5

The Gap is Wide Open

20+ companies build Operational RL environments. Zero target crypto. Wrapper-dependent agents hit capability ceilings on complex DeFi operations.

$4-8B

per year in RL environments

The Category is Proven

Operational RL is a validated, venture-backed category. The crypto vertical is untouched.

$2B+

Confirmed annual spend (Anthropic + OpenAI)

20+

Operational RL companies — zero in crypto

$1.3B

Applied Compute valuation in 8 months

Applied Compute ($1.3B), Mechanize, Surge AI validate Operational RL as a category
EVMbench: 70% bug detection with Codex — but detection is not execution
7+ platforms shipped agent Skills in one week — Skills abstract, BlockchainRL trains the capability beneath
Wing VC projects consolidation to 3-5 players by 2030. Vertical expansion expected 2027-28

Build With Us

We are looking for design partners who share the conviction that AI agents need native onchain training — not more wrappers.