The model can change. The knowledge remains.
Wingman is the first application built on Atelier — a persistent memory and reasoning platform that runs over the top of any model and any product, and makes it better.
Most software resets every time you close it. Most chat systems start over every conversation. Most tools tell you what happened and forget it by morning.
Nothing here replaces what you already run. Atelier runs over the top of a language model and improves it. Our own model is IAMRI V10, built to be more personable and cheaper to run, and it is the main product — but it can just as well be any model you already pay for; bring your own API key, because the model is not where the power is. It runs over the top of a product the same way: an exchange, a CRM, a service company's back office, a piece of legacy software nobody dares touch. We have taken companies' legacy systems, analysed what they actually did, and rebuilt them from that — the business kept its history and lost its limits.
A language model provides answers. Atelier provides memory, continuity, lessons learned, permissions, collaboration and accountability.
Models can be swapped. Knowledge remains.
Why memory matters
Without memory, mistakes repeat, research repeats, investigations repeat, and the same lesson is lost every time. With memory, previous work is reused, previous mistakes are avoided and previous discoveries stay available. Knowledge compounds. That is the entire reason a platform built this way gets more useful the longer it runs — without a new model being built.
The memory design is called KWSYF — Knowledge Will Set You Free — and it is not new. Mark Miles began it in 1994 and ran his own companies on it from the 1990s to 2014: a service business, a manufacturing business, a retail business and international shipping, in more than one country and more than one language. The surviving customer file from those years still holds 41,357 customer records and 34,490 orders. It allows work to continue where it stopped instead of being recreated from scratch, and it remembers the way a person does: ask about a day and the whole period around it comes back, not one isolated fact.
How the work gets done
Inside Atelier, different participants are responsible for research, review, planning, validation and execution. No single participant does everything, and a decision is reviewed from more than one perspective before action is taken. Before any action occurs the system already knows who can see, who can recommend, who can approve and who can execute — and every step is logged with who did it. When a question needs a tool the system does not have, it writes one and runs it.
What is learned is kept. A bug, an idea, a customer's comment lands in one inbox, is weighed, and either becomes work or is recorded with the reason it did not. Nothing begins from a blank page.
Public benchmark
Anyone can measure it. Run the benchmark with your own questions → It puts the same question to Atelier and to the leading frontier models at the same time and shows every answer in full — timed, tokens counted, every cited link checked, the real cost of each. Results, method and questions are in front of you. As Richard Feynman put it: “The first principle is that you must not fool yourself.”
It fits any business
Wingman itself: protection that follows the account, forensic reconciliation to the cent, opportunity discovery, and a platform that improves for every trader at once. See the trading product on this site.
Runs over the top of the system you already use. Customer history, routes, invoices, notes — remembered as context, answered in plain words, never re-keyed.
We read what the old system actually did — its tables, its rules, its twenty years of records — and build the new one from that. The history survives. The limits do not.
Bring your model or use ours. Add memory, lessons, permissions and a workspace to what you have already built, and it starts compounding from the first day.
Knowledge will set you free
The most valuable asset in any system is not data. It is understanding. Preserve it, build on it, never start over — so that every future decision benefits from everything previously learned.