FOR IMMEDIATE RELEASE
We measure systemic stress before crisis is obvious. The company that ships that instrument runs AI-native — not AI-in-the-number.
Entropy Index today published AI runs the company. Not the score.: what Maxwell, Lloyd, and the Cursor engineering loop actually do for a two-person founding core — and what they are forbidden to touch.
Who Lloyd and Maxwell are
Lloyd — Lloyd is Taylor's personal assistant: daily CEO brief, system status, and draft inbox under founder GO / EDIT / KILL.
Maxwell — Maxwell is Chief of Staff — the always-on operating layer: readiness, the progress loop, materials prep, platform hygiene, and the single approval inbox. Proposes; does not send or write the Index. Often 18 agents run underneath that seat.
Together they keep the operating cadence tight so judgment stays with the people who built the Index.
The split that matters
Most “AI risk” products put a language model in the readout. We put agents in the company, and keep the instrument sealed.
Data in. Score out — zero AI touching the engine.
One sealed formula on the live Index and every backtest — formulaic result, not narrative. Scoring is deterministic under sealed physics. Zero AI touching the engine: AI may structure evidence and read the face; it does not write the number.
That boundary is public doctrine on /news/sealed-boundary. Company operating agents are not that claim — and they do not rewrite it.
What is automated (honest share)
On a 100-point company system, the OS (Lloyd + Maxwell) credits about 58% — briefs, status, drafts, readiness, materials prep, platform hygiene, and half of agentic build prep. Send, capital, shipping the live Index, and live rooms stay human. GO is not send.
On that 100-point company system, the OS covers:
- Daily CEO brief and system visibility
- Draft inbox and pipeline hygiene
- Readiness / gate reporting and the autonomous ops progress loop
- Materials readiness — diligence packs, memos, attach packs
- Platform hygiene — CI scorecards, cron health, collector truth
- Half of agentic build prep — specs and bounded Finish queues (not Cursor writing physics)
- Shared preparation on public copy, bizdev priority, and GO / EDIT / KILL — founders still decide
Parallel research OS: Hermes is Travis's research-division OS. It credits about 44% of a 100-point historian-author system — manuscript delivery, The Insecurity Index lab commits, and framework hands from Telegram — separate from Maxwell. Physics, claim ceiling, and IC rooms stay Travis. He does not need a laptop for science delivery.
The loops run on a cadence so judgment stays with the people who built the Index.
What stays human
- Ship the live Index — product, evidence systems, entity architecture
- Live rooms and GTM — humans in the room
- Ambition, capital, partnerships
- Send — outreach does not leave the building without founder approval
GO is not send. Drafts are not published claims. Travis owns framework science; Maxwell tracks blockers and does not resolve them.
Why this is rare in risk intelligence
An AI-native risk platform is not a chatbot bolted onto a dashboard. It is continuous sealed scoring plus an operating model that compounds delivery without a services org — while refusing generative scoring in the number.
Desks and agents meet the same freeze: Watch for notice, Act for capital under dual Crisis, receipt and physics hash on paid reads. Distribution on x402 is the sealed instrument on the rail — not a swarm of autonomous scoring bots.
What we will not claim in this statement
- Not “AI predicts crises”
- Not “Maxwell writes the Index”
- Not organic GMV, licensed desks, or stranger-wallet PMF
- Not that automation replaces founder judgment on capital or send
### About Entropy Index
Entropy Index measures systemic stress before crisis is obvious — the same sealed instrument for desks and agents. Not a trade signal.
Media: Newsroom · Team: /about · Instrument freeze: /news/v5-lock