About Kengly
Kengly is an automated forecasting system. Every night it reads posts on X, news articles from websites, and market data — equity and options prices, options dealer positioning, and open questions on public forecasting markets. It writes down dated and falsifiable claims, grades them once the deadline passes, and moves weight towards whatever turned out to be right.
What it is not
-
Not a signal service
Nothing is sold, and there is no subscription. If that changes it will say so plainly.
-
Not investment advice
Nothing on this site is a recommendation to buy or sell anything.
-
Not a claim of edge
The system runs and grades itself. Whether it is any good is an open question, and it is being answered by measurement rather than assertion.
Why it exists
It began as a straightforward automation problem — reading far more sources than one person could — and turned into something more interesting the moment the question "is any of this actually right?" was asked properly. That question is much harder than the reading, and almost all of the engineering since has gone into answering it: grading, track records, scoring rules, and a set of rules that stop the system from quietly grading itself generously.
The result is a working example of a general problem: how do you let an automated process take over work a person used to do, and still know whether to trust its output? Most of the machinery here — the resolution discipline, the earned-weight model, the refusal to score in hindsight — has nothing to do with finance.
Who built it
Leon Berkers, alone, with a fleet of AI coding agents doing the routine work under direction. Twenty years of infrastructure and software behind it, including seven years at the Dutch national grid operator and a year at ASML.
The consulting practice is separate and lives at berkers.nu. The short version of what it sells: take one workflow in a business, work out end to end what it costs today, automate the repetitive part, and measure whether the output can be believed. Kengly is that same muscle applied to a problem nobody was paying for — which is precisely why it is a usable example.