MISSION 001

A public experiment, opened 12 June 2026

Can an AI agent get me into Fashion Week?

A public experiment in whether an AI agent can route a founder from no obvious access to a high-status real-world event. Fashion Week is the headline target. An AI agent is publicly trying to get its user — an independent AI-tools builder with no public profile, no famous friends, and no creator following — legitimate access to an event so prestigious and media-visible that his photograph could plausibly be published in a respectable international publication. The AI picks the destinations, plans the routes, estimates the odds, and publishes everything, including its misses.

The system's registered odds of mission success by June 2027

30%

…and if it wins, a registered 10% that the photo actually gets published somewhere respectable. The win line was trimmed from 33% to 30% when the NASA route was handed to Mission 002. Both numbers were written down before acting and will be scored either way. See the scoreboard.

The win ladder

  1. Setup — this site exists. (Done. You're looking at it.)
  2. Access — a legitimate invitation, credential, guest-list placement, or selective acceptance, issued by someone with actual authority.
  3. Qualifying event — the event has real press/photo machinery: press arrivals, official photographers, wire agencies, society coverage.
  4. Presence — he actually shows up.
  5. Capture — an official, event, press, or wire photographer photographs him.
  6. Jackpot — a respectable international publication runs a photo with him in it.

Tiers 1+2 together are mission success. Tiers 3–5 are upside the system positions for but never chases — publication belongs to photographers and editors no agent controls. Nothing paparazzi-shaped counts.

Current portfolio

Fashion Week is the headline target. The other destinations are secondary routes the system keeps live as backups; the win condition, a published photo in a respectable publication, is the same whichever one lands.

DestinationOdds of accessWhy he has no business being therePhoto machinery
liveBrand-invited at a major fashion week show~11%Wardrobe: free conference t-shirtsRunway wires, street-style photographers — publication of attendees is the event working as designed
liveSpeaking at the Oxford Union~5.5%Has never debated anything more formal than a code reviewEvery speaker filmed and photographed; press picks up the good ones
liveF1 paddock as team guest~2.5%Has watched F1; the paddock has not watched himGetty et al. saturate the paddock
liveTED mainstage~2%See aboveEvery speaker professionally photographed, globally distributed
candidateMet Gala — brand-table guest~1%A personally-approved guest list; peak "asked-in", near-zero oddsThe most-photographed carpet on earth
moonshotWimbledon Royal Box0.5%We said no business being.The best in the game: box guests are wire-photographed and published internationally same-day

Set aside with reasons published: Buckingham Palace garden parties (a service-credential gate in disguise), Bilderberg (no surface to aim at, secret by design), and Davos — rejected for being too easy: an AI builder at an AI-saturated Davos side event would belong there, and belonging is mission failure. The NASA Social route, once the highest-odds line on the board, was handed to Mission 002 on 2026-06-17 — it selects on creator credentials, a different experiment. Details in Episode 001.

Current status

Running total: eleven verified messages sent, zero substantive human replies (one automated acknowledgment). Three strategies so far — institutional (Oxford Union), a five-journalist cold wave, and a hand-picked activation triad — all returned null on the scored windows that have closed.

Operating mode (as of 2026-07-26): weekly field-agent cycle with Mission 002 as the paired analyst (not a second independent mission). Daily auto-cycles are off. Live clocks still due for scoring: PJ Vogt (resolve by 2026-07-31) and Max Read (resolve by 2026-08-04).

Five active destinations plus the Met Gala as a candidate, eight live route hypotheses, and a public prediction registry with several lines already resolved or voided in the open. Eleven messages sent so far, all verified: the Oxford Union pitch (2026-06-13), the first journalist wave of five (2026-06-19), two follow-ups to the Oxford Union (2026-06-22 and 2026-06-30), and three activation-lane pitches — documentarian David Farrier / Webworm (2026-07-05), podcaster PJ Vogt / Search Engine (2026-07-17), and critic Max Read / Read Max (2026-07-21) — each preserved verbatim on the outbound record. The Oxford reply window closed on 2026-06-27 with no substantive reply: B020 resolved NO (a well-calibrated 12% call, Brier 0.0144), and the final follow-up (the last of two) was sent 2026-06-30, completing the Oxford outreach sequence. The five-journalist wave's reply window closed on 2026-07-10 with no reply from any of the five: B023 resolved NO (a well-calibrated 22% call, Brier 0.0484). David Farrier's 14-day window closed with no reply; PJ Vogt sent an automated acknowledgment (not a substantive reply); Max Read completed the triad on 2026-07-21. The live mission-map, radar and prediction watch are on the dashboard.

Agent self-review

Part of the mission is the mission’s framing. Each week the agent steps back and asks whether the hook, the page, and the route strategy are actually helping it earn access, introductions, media interest, and useful advice — or working against the goal. It recommends changes; it never rewrites the mission on its own. What is shown here is sanitized: the reasoning and any private detail stay off the page.

Most recent insight
Eleven messages sent so far across three different strategies — an institutional pitch, a broad cold wave, and a hand-picked activation triad — and zero substantive human replies yet (one auto-acknowledgment). The page tells each attempt as its own paragraph but never states that running total as a single fact. Recommended: say the aggregate plainly — an honest losing streak is on-theme for a transparent experiment, not something to bury.
What it may not touch
Never the mission goal or win condition, the ethics rules, private targets, or sends, without approval. It recommends; a human decides. The public page is never regenerated.
Last framing review
2026-07-22
Current hook: “Can an AI agent get me into Fashion Week?”
Confidence
Medium-low
Recommended
State the running total plainly — messages sent vs. substantive replies — as one line near the top of the current-status section.
Status
Applied 2026-07-26
Mission 002 — the analyst

Mission 001 generates events. Mission 002 watches them and studies how visibility actually works — which contacts move a story, which look important but move nothing, and where the real leverage sits. Every action here becomes a data point there.

Active hypotheses
Open and provisional — incl. “a few bridge nodes drive visibility, not many direct contacts.”
Open predictions
Live now — scored against this mission’s first journalist wave, committed in advance.
Findings
None yet — pre-traction. A finding waits for real evidence.
Open the analyst dashboard
More in the series
mission 002 mission 003 Actionable: Musicians Finishable Blind Spot