Direct operating insight
You understand a construction decision deeply enough to identify the buyer, current behavior, failure modes, and evidence source.
Construction & physical-industry AI · Built through Good Combinator
Build a construction AI company around evidence.
Bring a narrow workflow, direct workflow insight, and the ability to ship against field behavior.
Founder fit
Who should apply to the 12-week build sprint?
Strong applicants combine construction-domain access with a narrow, pilotable AI wedge.
A real review loop
You can reach users, workflow owners, or pilot partners who will review shipped work rather than react to a generic demo.
One testable constraint
The first product surface is narrow enough to test within 12 weeks and important enough to change an operating decision.
Ability to ship every week
The team can produce reviewable progress, absorb field feedback, and document what the evidence does and does not prove.
Sprint sequence
Twelve weeks from thesis to reviewable evidence.
The sprint is organized around decisions and proof rather than feature volume.
- Gate 0
Fit review
Confirm direct workflow insight, buyer access, data availability, founder-market fit, and whether one pilot can be tested with real users.
- Weeks 1–2
Pressure-test the wedge
Choose one workflow, define the acceptance criteria, document the baseline, and remove features that do not help the pilot decision.
- Weeks 3–8
Ship against field behavior
Review product usage, source records, integration friction, user feedback, security constraints, and human approval boundaries every week.
- Weeks 9–12
Make the next decision legible
Package the build, measured evidence, limitations, deployment assumptions, and partner story into a pilot-ready evidence package.
What leaves the sprint
A clearer product and a more honest diligence story.
Expected outputs include a working product surface, documented workflow and data assumptions, a pilot plan, security and human-review boundaries, measurement definitions, and the evidence package needed for the next decision.
Next decision
Bring the wedge, buyer, and proof target.
Applications are evaluated on workflow insight, access, evidence quality, and weekly execution capacity.