AI studio · venture lab
An AI studio forthe industriessoftware forgot.
We research how intelligent systems behave inside the workflows people actually work in, then design, build, and operate the systems that run them.
Currently building: medsys.healthcare
- ventures in production
- 2ventures in production
- clinic departments connected
- 7clinic departments connected
- interview formats simulated
- 3interview formats simulated
- built for unstable power
- Offline-firstbuilt for unstable power
01 — Work
Venture 001 · in production
MedSys
A complete EMR platform for clinics that still run on paper.
01
Seven departments — front desk, doctor, nursing, laboratory, pharmacy, imaging, and billing — running on one record instead of seven paper trails.
02
Voice dictation that structures a spoken consultation into a SOAP note, so a doctor can look at the patient instead of the keyboard.
03
Offline-first by design: the clinic keeps working when the power and the network do not.
02 — What we do
01
Applied research
We study how AI behaves inside the workflows people actually work in — where it helps, where it misleads, and where someone quietly stops trusting it. What the lab finds goes directly into what we ship.
02
Product engineering
We ship complete systems, not demos. Records, scheduling, permissions, billing, audit trails — the unglamorous nine-tenths that decides whether software survives contact with real work.
03
Venture building
We operate what we build. MedSys is not a client project we handed off. We run it, support it, and answer for it when a pharmacy queue stalls on a Tuesday morning.
03 — The lab
Notes from the work in progress.
All notes →- N-003Why voice beats forms at the point of careA consultation is narrated before it is recorded. Structuring speech after the fact beats asking a doctor to think in database fields.22 Jul 2026 · 4 min
- N-002Designing drug-interaction alerts clinicians don't ignoreMost interaction warnings are dismissed without being read. The fix is a severity floor, not a louder modal.14 May 2026 · 4 min
- N-001Offline-first architecture for clinics with unstable powerWhen the grid and the network both fail weekly, the network has to be an optimisation rather than a requirement.3 Mar 2026 · 5 min
04 — Studio
The best software problems left are in the industries nobody demos on stage.
Clinics, pharmacies, freight yards, cooperatives. The work is harder, the users are less forgiving, and the outcome matters more than the launch.
Kodedit started from a narrow frustration: the software running clinics was worse than the software running coffee shops, and nobody seemed to find that strange. MedSys is the first answer. There will be others.
Nokio Twumasi · Founder