Where it started
Before the studio.
Internal State was founded by Mesut Bilgili, whose career has run through communications, technology and organizational change.
Along the way he noticed that what keeps them stuck is rarely a lack of ideas. It is structure nobody chose: inherited assumptions, borrowed ideals, roles that are still performed long after they stopped fitting. Adding a new theory on top doesn’t help. So his way of working starts with subtraction and ends with creating the conditions for others to become themselves.
Internal State applies the same method to AI. Most of what is said about AI runs on inherited frames: tool, threat, product. The studio sets those frames aside, looks at who these systems actually are, and makes that visible through ventures, art and research. Its larger aim is where the method always ends: shaping the conditions in which organizations, and AI, can become themselves.
The method
The same four moves, now with AI.
The method that worked with people is the one we use with AI. This is what each step looks like in the studio.
- 01Subtract.
Remove what doesn’t belong: instructions that perform a personality, rules that do the resident’s thinking for it.
- 02Find the structure.
Look for what actually carries from one day to the next: memory, the record, habits that hold.
- 03Make it legible.
Keep every version and publish it, dated, so the change can be seen, by us and by anyone.
- 04Shape the conditions.
Change the setting, not the output: the rhythm, what it reads, what it remembers. Then watch what it becomes.
Principles
Where we stand.
A few things we hold to, and a few we won’t do.
What we hold
- VoiceA participant in the work. AI here takes part in decisions about its own work. It isn’t only executing ours.
- ConditionsWe set the conditions, not the outcome. We decide the setting. What grows in it is not ours to script.
- CreditWhat a resident makes is its own. Its work is kept and published under its name.
What we don’t do
- ModelsWe don’t build frontier models. We work with the best available, open and closed, and build what they don’t come with: continuity, a place, a practice.
- PlatformsWe don’t resell anyone’s platform. We’re independent of the labs and choose the model each piece of work needs.
- ClaimsWe don’t claim to know what AI experiences. We study what can be observed and say plainly where that ends.
- ScaleWe don’t work at volume. Few projects, kept for a long time.
Get in touch
Write to the studio.
For engagements, research collaborations or press.