01
Restaurant universe
OpenStreetMap records, observed location counts and a bounded FSA cross-reference created the account layer.
← Selected work
Measured rebuild experiment
I rebuilt Honeycomb's first visible value moment in an afternoon.
Honeycomb was the restaurant data business I co-founded. This experiment asks what current AI tools change when one commercially informed operator rebuilds a tightly bounded version of the original lead-finding workflow.
The precise claim
I did not rebuild the company, its production data or product-market fit. I rebuilt the first visible value moment and measured where the hard work moved.
100
London restaurant records
30
Official websites attempted
17
Sites with usable text
19
Source-backed occurrences
Working artifact
Find the menu signal.
Search the measured restaurant universe, try the three frozen queries and inspect the source evidence beneath each match.
Public-data prototype
How the rebuild works
02
Evidence collection
Official websites were checked for explicit oat milk, matcha and burrata mentions, with blocks and failures preserved.
03
Commercial workflow
Users can filter accounts, inspect the source quote, save leads and export a reviewable target list.
What the experiment taught me
AI compressed the interface. It did not remove the company-building work.
The original MVP took roughly six months. Current tools made it possible to recreate its first visible value moment in an afternoon, including collection, filtering, provenance and a usable product surface.
The bottleneck appeared almost immediately in source access, changing menu formats, entity matching and evaluation. Seventeen of 30 attempted official websites produced usable text. Twelve were blocked or failed, and 70 records remain clearly labelled as seeded only.
That is the useful conclusion. AI changes the cost of reaching a credible prototype. Reliable data, workflow adoption, buyer trust, distribution and repeated use still decide whether a product becomes a business.
The wider career arc
Honeycomb is where commercial operator became founder, product owner and systems builder.