AMOngoing
Ambient Interfaces
What a product looks like when the interface assembles itself around the task instead of the menu.
ExperimentsRead the notes
Intelligility Labs
Where we experiment
first.
Labs is the R&D arm. New models, new interfaces, new agent patterns, and a lot of ideas that don't work. The ones that do end up in Studio, Apps or the Academy curriculum.
Most experiments fail. That's the point.
The tooling around AI development changes faster than any team can casually keep up with. So we don't casually keep up — we run structured experiments and write down what happened.
Some of it becomes product. Some becomes teaching material. Some becomes a short note explaining why an approach everyone is excited about doesn't survive contact with a real workload.
We publish the reasoning, not just the wins.
What this looks like
01
Interface research
What software looks like when the interface forms around the task instead of a fixed menu structure.
02
Agent architecture
Tools, memory, handoff and escalation patterns tested against workloads that actually have consequences.
03
Evaluation
Our internal harness for deciding whether a change made the product better or just different.
04
Speed benchmarks
How far a single builder can take an idea in one day, one week, one month — measured, not claimed.
05
Model comparison
Practical head-to-heads on the tasks we care about, with cost and latency in the same table as quality.
06
Failure notes
Short write-ups on approaches that looked promising and did not hold up. Useful more often than the wins.
Active experiments
A running list of what Labs is currently working through.
AGOngoing
Agent Handoff Protocol
A working pattern for when an agent should stop, summarize and give control back to a human.
ExperimentsRead the notes
ONOngoing
One-Week MVPs
A public log of ideas taken from prompt to deployed product inside five working days.
Internal ProductsRead the notes
LESelected work
Legacy Systems, New Leverage
Client builds where AI wraps around software that is twenty years old and not going anywhere.
Client BuildsRead the notes
EVOngoing
Evaluation Harness
Our internal scoring rig for deciding whether a model change actually improved the product.
Internal ProductsRead the notes
PROngoing
Prompt-to-Prototype
Teaching materials built from real builds, not toy examples. Feeds directly into the Academy.
AppsRead the notes
We don't just talk
about AI. We build with it.
If you have a problem that doesn't have an obvious answer yet, that's usually the interesting kind. Bring it to us.