The habit we already had
Around 2014, the firm that came before us encoded the per-trade numbers its subcontractors price from, so it could stop waiting three weeks on trade bids before telling a client what a job would cost. The habit worked. It never became a product.
By 2026 LÏEF Development was meeting the old problem again: outside takeoff vendors that were slow, and rate sheets scattered across folders. We kept running into the same wall. What we lacked was not an idea. It was someone to build the engine to the standard a builder could rely on while the estimating seat kept estimating.
Common Ground filled that gap. Jesse Fowler, who holds the estimating seat at LÏEF, was the person from Common Ground who facilitated the build.
What the work looked like
April 2026. The pipeline took a plan set to a priced bid with nobody at the keyboard. The first autonomous job, a 6,919 SF premium spec home, finished in 3 minutes 31 seconds.
May 2026. A quick-service restaurant bid went out with its markup layers compounding on one another. It was caught, fixed in the second version, and the check now fails the build by itself. We counted that as a good result: a mistake that cannot repeat.
The audit. Before we said anything in public about accuracy, we had a frontier model go looking for every place the engine marked its own homework. It found 75 of the 167 catalog rates copied from one anchor project's budget. The flattering number did not go out.
A stranger's test. A VP of Operations at a building-materials manufacturer ran the engine on his own plans. A blind second pass, run without sight of the first, caught a wall-height error in our own first pass that would have shrunk a shell by about a third. He asked to buy the method. We kept it, and the blind pass is now required on every takeoff.
September 2026. The front end stopped reading a picture of a drawing and began measuring the drawing's own geometry, about 49,000 paths on one sheet. A person inspects an overlay before any quantity is trusted.
The rule we publish by
Results are public. The method stays private. And we claim only what was earned on bids the engine had not seen. Today that is one figure: across three single-family bids priced blind, a mean error of 18.6 percent and a worst bid of 31.9 percent, set June 10, 2026 and re-run as the corpus grows.
We describe the way we estimate the same way our site prints it: Kanopi measures each plan set twice, the second pass blind, puts a person over the overlay, and calibrates to a client's own history before a live bid ships.
Being found the way people ask now
A builder deciding whether to trust an AI estimator will ask an assistant, or search, and read what turns up. We want that to include our method pages and our proof, not a competitor's comparison table. So the Common Ground wiki holds dated pages on how the engine works and the questions people ask, written to be read by a person and cited by a model. Our own site's three pages cover the platform, the proof and the start of a project.
Where it stands
Eight builders and developers now price their own work through Kanopi, as of September 25, 2026. A full takeoff that took about 40 hours takes about 40 minutes, delivered in 24 to 48 hours with the blind pass and the overlay included. The first bid is free, in the client's own format.
If a builder asked us whether to work with Common Ground, we would say: start with what you will not claim.
Credits
- Common Ground VenturesCreated by Common Ground Ventures. The engine, its checks, its public proof and the website.
- Jesse FowlerPrincipal, LÏEF Development, in the estimating seat
Related Common Ground Ventures work
Kanopi's problem is a measurement that has to survive being checked by someone who is not on your side. Common Ground's work in other industries meets the same problem: a deep tech method that needs a buyer's trust, an evidence-graded reference, and a concept that has to say plainly what stage it is at.
On the record
Common Ground Ventures built Kanopi's engine, its checks and this website, and helped write this wiki.