Notes from building and running our own software: what breaks in estimating, why we calculate rather than guess, and what a venture studio does differently. Plus practical pieces for the trades we work with.
Estimating still runs on spreadsheets and memory. The failure modes, why they compound on margin, and why deterministic take-off is the structural fix.
Vision AI reads the drawing. A deterministic rules engine calculates the quantities. Why that boundary decides whether a quote can be defended line by line.
A take-off derives every material quantity from the geometry of a plan, not from a list someone types, and here is where software keeps dropping the chain.
When a builder challenges a line on your quote, the answer should already exist: the rule that produced it, the version it sits in, and who changed it.
A venture studio starts its own companies, builds the product itself and keeps operating it: how LoopSmiths picks, builds and runs its ventures.
When the team that writes the code is the team that answers for it, the design changes: explicit rules, a record built in, and failures that are loud.
Property software holds records well but does little to help produce them, and producing them is where inspections, chasing and disputes eat the week.
A missed call is rarely a message waiting for you. It is a lead ringing the next number, and that is a mechanism you can do something about.
Quotes rarely go cold because the price was wrong. What silence usually means, and how to build follow-up that holds a quote open without annoying anyone.
Most tradies know reviews matter and still do not collect them. Here is the timing, the channel and the step that makes asking feel normal.
One venture live and trading, the next in development.
