Why Thinkery exists.
We organize the knowledge machines work with — open, verifiable, under human control.
A letter from the founder
01 The letter
Thinkery started as an experiment. I wanted to know how far a single developer can get when he takes AI tools seriously. The answer: astonishingly far. But the longer I worked with these machines, the more often I found myself explaining to them in the morning what they had already known the night before.
That is everyday life with AI: no memory, no provenance. What a machine learns today is lost tomorrow. Where its knowledge comes from, nobody can say. And what it was allowed to know, rarely anyone decided deliberately.
Yet for everything people care about, they have built systems of record: bookkeeping for the money, archives for the contracts, registers for the property. Only the knowledge machines work with has nothing of the kind. At first I found that astonishing. Then I thought: apparently nobody is going to build this for me. So I started.
I believe in efficiency because it doesn't belong to big corporations alone. AI gives far more people the chance to realize their ideas: the craft business, the teacher, the solo founder. But that only holds if knowledge is allowed to accumulate instead of being rebuilt every morning.
And I believe in a second thing that matters more to me: control. People should determine what an AI reads, stores and recalls — not the companies behind the models. Trust must be checkable, and it must never be for sale.
That is why Thinkery exists — a lab, not a classic software house. The name is a wink: think plus -ery, like bakery — a place where thinking happens. We experiment, throw a lot away, and what works becomes a product. The first field forced itself on us: giving the knowledge of machines its rules. The craft is called context engineering, the responsibility context governance — terms that are only now emerging. We are helping to shape them.
We write the open standard and deliberately hand it over. We build the tools that bring it to life — from the open-source engine to the platform for entire organizations. And we guide everyone who wants to put their knowledge in order before their machines do.
What drives me is curiosity. I try things, run experiments, throw a lot away — and the best part is not doing it alone. The core of our work is open source, the specification grows through a public process, and some of the best ideas so far came from people I have never met.
If this speaks to you: build along. Or just write to me — I answer myself.
Yves Gugger
Founder, Thinkery
02 The discipline
Context engineering & governance.
Behind the thinking layer sits a young craft with two halves. We work on both — with the open standard, the tools and work done in public.
The craft
Context engineering means preparing knowledge so machines can work with it. Selecting, distilling, versioning — engineering work, not a prompt trick.
The responsibility
Context governance means ruling and proving what an AI may know and did know. Who approved it? Where does it come from? What did the machine see when it decided?
03 The values
Three properties, not slogans.
They are in the mission because they are built into the products. Each one can be checked — and each one has a price we pay deliberately.
-
Open
Every interface is open, every layer replaceable. Using one of our products never forces you to buy a second one. The standard belongs to no one — including us.
ctxpkg.org never monetized · public RFC process · two independent verifiers
-
Verifiable
What an AI knew can be proven — cryptographically signed, checkable offline, without asking us. Trust comes from mathematics, not marketing.
Ed25519 signatures · trust reports · evidence bundles
-
In your hands
People determine what an AI reads, stores and recalls. Nothing leaves your machine without your command — and trust is never for sale: rankings and verification cannot be bought.
Zero telemetry · local-first · policy packs · no paid placement
Where this leads
A thinking layer in every organization.
As ordinary as bookkeeping is today: a place where knowledge stays, is proven, and people decide. That is what we measure ourselves against — in years, not quarters.
Build along — or write to us.
The core is open source, the standard grows in public, the door is open.