01draftRead
Fixed rules, thinking colleagues
Fixed code wherever possible, and language models where they fit
ITARR Lab
AI will affect every business in the Netherlands. ITARR builds working AI solutions for small and medium-sized businesses and researches the leading edge of the field. We share what we learn.
Start at step 01 Replay the night
Only your own browser remembers what you have read.
Point to a step or tap it.
01draftread
Fixed rules, thinking colleagues
Fixed code wherever possible, and language models where they fit
ExplanationControl5 min read
02read
Systems asking for room
What happens when AI tasks share memory and computing power
ExperimentOwn ground
Planned
03read
Custom models on your own ground
Where each task runs and what it costs
Working paperOwn ground
Planned
04read
A person at the helm
How a person authorises irreversible actions
ExplanationControl
Planned
05draftread
A night in the workshop
What happens when two systems keep triggering each other
ExperimentWorkshop±3 min play
06draftread
How does an AI colleague catch its own mistakes?
Seven ways we observed in our workshop in one week
ExplanationWorkshop5 min read
07read
One face, two jobs
How our AI chooses where to think
Working paperOwn ground
Planned
08draftread
Stay in control of your AI
A working policy for AI that does real work
Working paperControl10 min read
Start with part 1. Then you will read the cases differently.
Part 1
The foundations: fixed rules, thinking colleagues and where our work runs.
01draftRead
Fixed code wherever possible, and language models where they fit
02Planned
What happens when AI tasks share memory and computing power
03Planned
Where each task runs and what it costs
04Planned
How a person authorises irreversible actions
Part 2
Real incidents from our workshop and what we learned.
05draftRead
What happens when two systems keep triggering each other
06draftRead
Seven ways we observed in our workshop in one week
07Planned
How our AI chooses where to think
Part 3
The method and our own commitments.
08draftRead
A working policy for AI that does real work
09draftRead
Draft. This is not a fully confirmed general commitment. Commitments 5 to 8 still require confirmation and further detail.
This is a draft. These commitments have not yet been adopted. For items 5 tot en met 8 we still need to confirm that we can fulfil them.
Each publication belongs to a field, shown as a label.
| Date | Field | Type | Title |
|---|---|---|---|
| Workshop | Experiment | A night in the workshop draft | |
| Workshop | Explanation | How does an AI colleague catch its own mistakes? draft | |
| Control | Explanation | Fixed rules, thinking colleagues draft | |
| Control | Working paper | Stay in control of your AI draft |
Adapt it for your business and tell us what you learned.
Do you work at a school, college or university? Use our publications and join the conversation.
Want to do this work in practice? Get in touch.