Le processus de recherche d'un fonds spéculatif, mené par des agents IA

A fund splits the work across specialists rather than asking one analyst for a verdict. Sector and fundamental analysts cover the business, a valuation team works out what it is worth, and a positioning desk reads filings and flow. The findings are argued — someone makes the bull case, someone makes the bear case — and a research head rules on which argument the evidence supports. A risk committee stress-tests what survives before a portfolio manager makes the final call.

No. Valarn is an educational research tool. We do not manage money, hold client assets, place trades, or execute orders, and we are not a registered investment adviser or broker-dealer. We simulate the research process a fund uses; everyone gets the same process on whichever stock they choose.

The process is reproducible even though the resources are not. Valarn runs up to 25 AI agents through the same sequence a fund research desk follows — specialist analysis, an adversarial bull-vs-bear debate, a risk review, and a single research view — in minutes, on any US-listed company. What a fund has that Valarn does not is order flow, direct access to management, proprietary datasets, and capital.

Two agents argue opposite sides of the same evidence over several rounds. A third agent, the Stock Reviewer, then rules on which case the evidence actually supports. The point is that a thesis has to survive an argument before it reaches you, rather than being the first plausible answer a model produced.

No. Valarn produces a research view — a stance with the reasoning behind it — never a recommendation to buy, sell, or hold. Every report ships with its sources, the disagreements between agents, and a data-quality assessment, so you can see what the view rests on and judge it yourself.