Le Sottisier

Every trade is proposed by one model voice — Bouvard — and cross-examined by another — Pécuchet — before any of it reaches a subscriber. The brief you receive is not a thesis. It is the residue of an argument: what survived being attacked from the opposite direction with cited evidence, plus the specific observables that would invalidate each position ahead of its formal stop.

The publication began on 2026-03-25 as Shorting SaaS — a single-thesis newsletter built around the claim that AI is about to gut per-seat SaaS valuations. It rebranded to Le Sottisier on 2026-04-25 alongside the promotion to a multi-model adversarial format. The new name is honest about the new method: a catalogue of received market ideas held up to scrutiny.

Bouvard runs the proposing model. Each market day it generates trade setups — directional, pairs, options structures, hedges — with entries, stops, targets, and a stated rationale grounded in the day's data: market microstructure, sentiment regime, supply-chain state, insider flow, macro context.

Pécuchet runs the challenging model. It does not score Bouvard's work; it attacks it. Every proposed trade gets a structured rebuttal that cites specific evidence — a contradicting print in the tape, a regime indicator pointing the other way, a missed correlation, a stale assumption. Where the rebuttal lands, the trade is killed before it ships. Where Bouvard answers the objection convincingly, the trade survives with a revised conviction score that reflects the cross-examination, not the initial pitch.

Subscribers see only what made it through, accompanied by the retraction triggers — concrete observables that would invalidate each surviving trade ahead of its stop. If the trigger fires, the position is publicly retracted before the market forces the issue.

The system tracks a broad universe across equities, commodity futures, currency pairs, fixed income, and sector indices, ingesting from market feeds, regulatory filings, news intelligence, social sentiment, insider transaction records, and macro indicators. Three analytic layers run in parallel: a structure-detection layer that finds patterns nobody asked about, a continuously retraining ML model that maps event-impact profiles, and a prediction-scoring layer that grades the system against what actually happened and redirects search priority where blind spots show up.

Bouvard and Pécuchet sit on top. Both consume the same upstream signal; they disagree about what it means. The disagreement is the product.

Flaubert spent the last years of his life on Bouvard et Pécuchet, an unfinished novel about two retired clerks who methodically work through every domain of human knowledge — agriculture, medicine, philosophy, theology — and find each one wanting. The companion volume he planned alongside it, Le Sottisier, was a catalogue of the received ideas his clerks had spent the novel demolishing: confident statements that fall apart on inspection.

The publication takes its name and tone from that project. Most market commentary is sottise — fluent, plausible, and structurally unwilling to be wrong. The point of the cross-examination is to ship the part that isn't.

This platform is for informational and educational purposes only. Nothing published here constitutes investment advice, a recommendation, or a solicitation to buy or sell any security. All investment decisions should be made with the guidance of a qualified financial advisor. Past performance of any thesis, model, or strategy does not guarantee future results.