It is intelligence quietly added to the things we already do. I care about what it notices, what it misses, where humans still matter, and whether it actually helps.
a working belief, not a manifesto ↗I may be underestimating how much the interface itself matters.
Experiment #001
Something that actually worked
built + tested
words ↓ signals
WORK INTELLIGENCE · 2026
What if meeting transcripts could tell me how every project is actually doing?
I built an intelligence layer over project meeting transcripts. It accurately summarized what teams were working on and surfaced blockers, action items, project summaries, and risks.
watch the conversation become something useful ↓
MEETING TRANSCRIPT
09:04Priya: The ingestion workflow is complete. We are validating the edge cases now.
09:11Dan: The source-data mapping is still waiting on the client decision.
09:26Ravi: I’ll send the validation sample before Thursday’s review.
09:39Priya: If mapping approval slips a week, the testing window gets squeezed.
Four moments. Useful, but easy to lose.
HOW THE EXPERIMENT CHANGED01 / 05
Question
Could a transcript reveal how a project is actually doing?
The hypothesis was simple: meetings contain more project truth than status reports capture.
The first version confused activity with progress. That was a useful failure.