Thinkery · Wädenswil, Switzerland

Make AI work better.

Thinkery is an applied AI company focused on context and efficiency. We work with companies that already use AI — improving workflows, making company knowledge usable, and building specialized agents.

I’m Yves, founder of Thinkery. I work directly with clients while building the team around this work.

Portrait of Yves Gugger, founder of Thinkery Yves Gugger · Founder

The hard part is context.

The model is rarely the thing that is failing.

Once AI is already being used, the difficult questions move beyond the model itself. What does the system know about the task? What happened before? Which information matters now? What may it do? What can be reused instead of reconstructed?

Companies already hold most of the knowledge their AI needs. The challenge is turning the right part into usable context without pushing everything through every model call.

That is the area Thinkery is going deep on.

From question to system.

We do not start with a predetermined AI deliverable. We start with the work and a question worth answering.

We understand the workflow, form a hypothesis, build enough to test it and measure what changed. Depending on the problem, that can stop at an experiment, become a production system, or be built together with the internal team.

The same people stay close to the problem from the first conversation through implementation. Strategy, design and engineering are not handed from one department to another.

01
Understand The workflow, the systems and where the real friction sits.
02
Hypothesis What we think should change, stated so it can be wrong.
03
Build Enough of the system to test the hypothesis properly.
04
Measure What actually changed, against what we agreed to watch.

Latest Thinkings.

The questions behind the work, published while we are still working on them.

What an answer really costs Aug 2026 · Efficiency · Read →

Every answer carries three prices: francs, seconds and watts. All three move with the same number, and the third appears in no dashboard.

Context is a budget Aug 2026 · Efficiency · Read →

Tokens behave like money. Unbudgeted, they get spent without anyone noticing. Why a longer window is not free, and how more context can make answers worse.

Retrieval is not search Jun 2026 · Retrieval · Read →

Search finds what is similar. Retrieval has to decide what a task needs. On embeddings, how documents get cut up, and the metric that misleads.

The cheapest iteration is the one in your head May 2026 · Method · Read →

Why we write the question down before we write code. With an example that did not work, and the reason it still sits on the shelf.

All Thinkings

What we’re building.

We also build our own technology around the problems we keep encountering.

LeanCTX

LeanCTX is our product for selecting, reusing and measuring AI context. It started from a problem in our own AI development workflow and grew into a system for reducing repeated context without treating quality as optional.

Experience from

  • GenTwo
  • Zühlke
  • Adnovum

LeanCTX · Starred by engineers at

  • Google
  • Microsoft
  • Amazon
  • IBM
  • Red Hat
  • Tencent
  • Autodesk
  • Dynatrace
  • Dell
  • VTEX
  • Zühlke

Engineers at these companies have starred LeanCTX on GitHub. That is a fact about the product, not a client list.

Bring us a question.

If your company is already using AI and you are trying to improve a workflow, make knowledge usable or build a specialized agent, tell us what you are working on.

hello@thinkery.ch