From Activation Sandbox to Training Layer

We returned to simulation with a different premise. Strategy League had asked users to practice investing and build public records. Activation Sandbox would preserve a real action the user already wanted to take.

Paper trading records virtual performance. An Activation Sandbox preserves real intent.

From practicing returns to preserving intent

Imagine a user who wants to deposit USDC into a Base lending market but is not ready to commit funds. Instead of joining a generic simulation, the user could rehearse that exact action with virtual funds, save it, monitor what would have happened, and activate it later.

The core output was a Shadow Action: a structured record of the target protocol, asset, amount, expected transaction, risk conditions, and activation rules. This was more useful than a fictional portfolio because it could become a real transaction when the user was ready.

A realistic wallet flow

The idea expanded from a web simulator into a training layer. Users could connect the same wallet, inspect the same contract calls, see the same approvals, and experience the same sequence without moving real assets. Protocols could use it to educate and activate cautious users. Agents could rehearse actions before receiving execution rights.

This reframing opened a larger market: onboarding, developer testing, agent evaluation, transaction education, and protocol activation. It also appeared to rescue the useful parts of simulation without requiring a public game.

Realism raised the cost

Every step toward realistic training introduced infrastructure work. Accurate state needed forks or simulations. Prices and rewards changed. Protocol flows differed. Wallet behavior, approvals, bridges, and failure cases had to be reproduced. A misleading rehearsal could be worse than none at all.

There was also a product tension. If the sandbox was simple, it would not prepare users for real execution. If it was realistic, it approached the cost and risk of building execution infrastructure—without completing the action that users ultimately wanted.

The missing evidence remained basic. Would users rehearse a DeFi action before doing it? Would protocols pay for simulated activation? Would a saved Shadow Action materially increase real conversion?

Training Layer was a more coherent use of simulation than Strategy League. But coherence was no longer enough. Before building realistic infrastructure, we needed proof that rehearsal itself was the behavior users wanted.