Craftpine
We build AI for the physical economy.
A Toronto research and development studio.
Programs are published in full, including the ones we close.
Programs
Corner
Concluded July 2026
A payments and inventory platform for independent retail: shopper checkout, a cash and card walk-in register, tap-to-pay terminal rails, and an unattended store that charges on exit. Built to production grade, taken to a live transaction on its own rails, and closed on the evidence of the retail category it was built for.
Alcove
Publishing
The practice that publishes the work. Holds the 520 Barton feasibility study: nine chapters and four appendices covering the category’s public record, the regulatory position, the machine survey, the unit economics, and the verdict that closed Corner’s retail format.
The record
Corner, built and closed between April and July 2026. Counts are audited at the close-out commit, 2026-07-14.
- Build window
- 76 days, 2026-04-30 to 2026-07-14
- Commits
- 173, one principal
- API routes
- 57
- Tables
- 20, row-level security on every one
- Migrations
- 48
- Regression suites
- 44
- Money paths
- Four: shopper checkout, walk-in register, tap-to-pay terminal, unattended charge-on-exit
- Adversarial pass
- 19 scenarios, 71 checks, all held. One real defect found, fixed, re-verified
- Live transaction
- $1.99 charged and refunded on its own rails, 2026-07-09
- On-device
- A 3B language model and speech-to-text co-resident on a $250 board; English voice end to end at about 2.6 seconds against a 4-second bar, with one of three languages tested over it
- Venue capital
- Approximately zero
It closed because the cost to prototype, develop, market and sell the retail format exceeded the return available at any demand anyone has measured. Roughly $26,000 of capital pays back in 2.2 to 3.3 years against a 24-month kill line. The only throughput figure the category has published is about 6.8 cups a day, from a mature unit in a captive venue, against a floor case that needs 12. Venue placement runs through incumbent vending contractors. Better-funded entrants ran the experiment first and lost: Briggo raised about US$19 million and was acquired by a Coca-Cola subsidiary, its format since wound down; Bobacino raised about US$3.5 million and closed without advancing past prototypes, its closure statement citing the difficulty of securing resources for development and regulatory compliance. The idea was vetted thoroughly and did not survive the vetting.
What was never proved is the market. No keeper was onboarded, no shopper used the app in the wild, no cup was counted. The platform was never falsified as a business; the venues for testing it were.
The full study, including the kill criteria set before any capital was committed: alcoveretail.com/studies/520-barton
Research and development
Experimental development: advancing technology by resolving technical uncertainty through systematic investigation.
Craftpine Inc. is an Ontario corporation, incorporated 2026-05-15, NAICS 541715. The 2026 program is documented to that standard: 56 slice specifications and 77 report files at close-out, kill criteria written before capital, and primary-source verification of load-bearing claims.
Contact
For media, partnerships, and research collaboration.