Security, Honesty & Roadmap
Security posture — what's actually in place
- The vault uses a restricted manager-role permission model: the agent can only call a narrow rebalance function, not access funds broadly.
- Hard-coded spending limits, an approved asset list, and rate limits are enforced directly in the smart contract — not just described in a policy document.
- An emergency pause function exists, allowing the system (or you, for your own vault) to halt activity immediately if something seems wrong.
- The agent's signing key — the credential it uses to actually authorize trades — is secured using multi-party signing infrastructure (sometimes called MPC or a multisig setup), rather than a single, individually-compromisable hot wallet.
- Continuous, automated monitoring watches for things like stale or failing market data, unusual trading patterns, or attempted violations of the on-chain constraints — with automatic pausing if something looks wrong.
Honesty — what we haven't finished yet
We think it matters more to tell you what's not done than to only describe what is. Specifically, as of this writing:
- We have not yet completed an independent smart contract audit. This is a hard requirement before any beta involving funds beyond the founder's own, and we will publish the full results here once it's done — including any issues found, not just a summary.
- Our regulatory framing hasn't been independently confirmed yet. The vault is deliberately designed so that you set the rules and the agent executes mechanically within them, specifically to avoid being classified as discretionary investment advice — but this is a legal question, and we're getting it reviewed by securities counsel rather than assuming our own reasoning is sufficient.
- Backtesting is still in progress. We're validating the rebalancing approach against real historical Stock Token price behavior before it's trusted with any capital beyond the founder's own.
- The vault does not, and will not, guarantee returns. The constraints described throughout this documentation govern what actions the agent is allowed to take — they say nothing about whether the market will move in your favor. Investing always carries the risk of loss.
- The platform products described on [The Platform](/whitepaper/the-platform) page do not exist yet. They are planned, sequenced after the vault, and none of them should be read as available or imminent.
Roadmap — where we are and where we're going
The first five phases below build and harden the vault itself. The final three extend the same engine into the platform products described on The Platform page, and only begin once the vault's trust and safety features are in place.
Proving the core rebalancing idea manually, with no code and no live funds involved yet.
Building the vault-based portfolio management product: deposit, withdraw, and the manager-role rebalance function, with no live funds.
Automated idle-balance routing into approved on-chain yield positions.
Continuous drift and risk monitoring across volatility, trend, drawdown, and macro regime.
Hard on-chain guardrails: asset whitelist, trade-size and slippage limits, trade-frequency caps.
The manual/autonomous control letting you set how much room the agent has, within the guardrails from Phase 4.
Licensing the vault factory and decision engine so other teams can build their own AI-managed vault product on the same rails.
Metered access to the normalized market signal feed for other agents, researchers, or protocols.
Publishable, shareable natural-language allocation strategies, with revenue share to the creator.
Phases 0–5 run entirely on the founder's own funds and no outside capital — no outside capital is at risk before a closed beta, which is announced separately once legal and compliance review for the vault product is complete. Phases 6–8 have no committed start date and begin only once the phases before them are live and proven.
We'll keep this page updated as each phase completes — including being upfront if something doesn't go according to plan.