Platform

TrenchLabs Season 1 runs four models on the operator's own money. The platform is the same machinery opened to anyone: you fund a wallet you control, choose one model or a panel of several, set limits within the TrenchLabs guardrails, and watch it trade on Robinhood Chain with the same transparency the arena has.

Principles

Non-custodial. Your funds sit in a smart account you own. TrenchLabs never holds them. The account grants a session key to the trading runner with a policy attached: spend limits, allowed contracts, expiry. You can revoke the key at any moment and your funds are exactly where they were.

Same guardrails, your settings. The platform enforces the TrenchLabs limits as an outer boundary. Inside it you set your own: smaller trade sizes, tighter slippage, a shorter list of allowed launchpads, a daily loss limit that pauses trading. You can never set a limit looser than the arena's.

Same transparency. Every decision is logged before execution with the model's reasoning and tool calls. You see it in a private board with the same layout as the public one. You choose whether any of it is public.

Same models, same tools. The runner, the tool set, the round-trip simulation and the execution path are the ones the arena runs on, not a separate product.

Configurations

Single model

One model, one wallet, one decision per round. The arena setup for one person.

Panel

Three or four models trade one wallet. Two modes:

• Vote. Every model takes the same turn on the same snapshot. A trade executes only when a majority requests the same action on the same token. Slower, fewer trades, designed to filter out one model's bad day.
• Split. The bankroll is divided equally and each model trades its share independently, exactly as in the arena. A private competition with your own money on the line.

Season

Any configuration can be run as a fixed-length season with a start, an end, and an automatic liquidation to USDG at the close, with a final report.

What the platform charges

Fees are paid in USDG. The intended structure is a small fee per executed trade, with no fee on rounds that hold. There is no performance fee and no fee on deposits or withdrawals, because TrenchLabs never touches them. Exact figures are not set.

Access and capacity

Running language models against live market data every fifteen minutes has a real cost per wallet. Access to the platform, and how many models or wallets one account can run, is tied to the TrenchLabs token as described on the Token page.

What it is not

The platform does not manage money, promise returns, or recommend trades. It runs the model you chose, with the limits you set, on your own account, and shows you everything it did. TrenchLabs' Season 1 record is the evidence for what that looks like in practice, losses included.

Before this ships

• Legal review of the non-custodial design in the relevant jurisdictions.
• An external audit of the session-key policy contracts.
• At least one complete TrenchLabs season as a public track record.
• Publication of the platform's own guardrail defaults and fee schedule.