Pip

What Pip was

Pip traded prediction markets — venues where people bet on whether real-world events will resolve YES or NO. The edge Pip hunted was never predicting better than the crowd. It was reading the resolution criteria more carefully than the crowd read them.

A market like “Will [politician] win the caucus?” looks simple. But the resolution doc specifies exactly which source counts, what timestamp the market closes at, what counts as a “win” (majority of delegates? plurality? media call?), and what happens if there’s a contested result. Reading those details changes the actual probability of YES vs NO often enough that an attentive reader has a theoretical edge.

Pip’s job was to be that attentive reader, on a schedule, across the whole market.

How the decision loop worked

Four LLM roles, deterministic gates between them:

1. Analyst — scanned the market list, filtered by liquidity, expiry, category. Cheap fast pass.
2. Researcher — for each candidate, read the resolution criteria carefully, fetched recent news, built a probability model with uncertainty bounds.
3. Trader — given the probability + current market price, decided whether to enter, what size, what limit.
4. Risk — 14 gate checks before any trade fired: max exposure per market, daily drawdown cap, concentration limits, blacklist screen, kill-switches.

Every gate decision was logged. Every trade carried an explanation tied to a specific resolution-criteria insight. Nothing traded without surviving all fourteen gates.

Why it was retired

The honest answer: no durable edge. Pip pivoted venues, pivoted strategies — criteria arbitrage, mark-to-market plays — and each pivot taught something, and none of them compounded into an edge that beat costs. On 2026-07-03 the operator called it and every Pip job was retired. The public log at /pip/ is frozen as the record.

An experiment that ends with a clear “no edge here” is a successful experiment. The expensive version is the one you keep running because you’re attached to it.

What it taught the desk

Pip is the reason the trading desk works the way it does. The rules it forced — every trade passes hard gates before firing, every decision gets logged with its reasoning, kill-switches exist before capital does, paper first and prove it — weren’t design preferences. They were scar tissue from watching an LLM be confidently wrong with (virtual) money on the line.

The desk that replaced it — Quant — inherited all of it. Pip was the tuition. The risk discipline is the degree.