Real work over benchmarks
Judge AI by what it can do on real, unfamiliar tasks, not only by a fixed test score.
03 / ABOUT
An AI + X neo lab for what comes after chatbots.
We build AI that takes goals into the real world, learns from the results, and improves with experience.
01Who we are
AI + X. AI, joined with a real field.
sagtuku is a neo lab: a small, research-first team that ships real systems. “X” is the field AI is paired with: software, science, engineering, and later industry and robotics. We don’t just study AI in the abstract. We put it to work in each field and learn from the results.
sagtuku is operated by Alpai GmbH, a Swiss company that turns generative AI into working systems.
02Our mission
Build the world’s most capable closed-loop intelligence system.
Our long-term goal: AI that can solve hard problems on its own, in software, research, engineering, industry, and, in time, the physical world.
03What makes us different
THE CATEGORY / CLOSED-LOOP INTELLIGENCE
From “train → deploy → answer”
to “learn → act → improve.”
Large AI models are great at reasoning and writing. We connect them to long-running tasks, real actions, independent checks, and memory, so they keep getting better.
LEARN→ACT→OBSERVE→VERIFY→REMEMBER→IMPROVE
04Research culture
How we work.
Evidence over stories
Treat every result as a claim to check. Make work repeatable and keep the evidence that could prove it wrong.
Build to learn
Put systems where actions have real effects, so we can see what works and what fails.
Small teams, clear ownership
Give researchers ownership of important problems and the freedom to build the answers.
05What we are building
One learning system.
Many kinds of tasks.
We are not building a pile of separate AI products. Everything we build strengthens one core system that can act, check, remember, and learn across many tasks and settings.
We start with technical research and engineering in digital settings. Robotics and physical-world work come later. They are a long-term goal, not our first step.
Read about our research↗WORK WITH US
Start with a hard question.
Tell us what your team wants to learn, build, or prove.