Research
Short papers from the lab.
- What tasteslop is made of
- What a polymath adds
- What lies under a feeling
- Build on a model, or train one
- Models write alike, not necessarily worse
- Advice from a machine that agrees
- What a model can learn about one person
- Weights, memory or a page
- Why designers use AI and still recoil
- What Fable 5.2 showed
- Fable 5.2 and the world-model question
- Who chooses what a world model builds
- What personal AI can cost
- What breakthroughs have in common
- Showing someone your world
- What other researchers found
- Forty concepts, three ideas
- Far is not new, and far costs fit
- A selector that learns one person
- What art survives automation
- The open loop
- Method, not motif
- What juries reward
- Two masters in one room
- The shape of a framework
- Sounding like him is not predicting him
- Closed loops converge
- A way of thinking travels
- A model judge is one voter
- A lean, not a sweet spot
- Taste is mostly personal
- One idea in thirty outfits
The results in one place: Findings.