Insights
What we’ve learned rebuilding how we build.
Notes from the delivery floor on AI-first engineering — what it genuinely changes, what it doesn't, and the failure modes nobody puts in the sales deck.
What a nine-week MVP has to leave out
A fixed timeline is a series of refusals. Here's what we cut, what we never cut, and how to tell the difference.
Read itEvals, not vibes: measuring an LLM feature before you ship it
Most AI features ship because a demo impressed someone. Here's what to build instead.
22 July 2026 · 10 min read
Engineering · ProcessHow we review code we didn't write
The instincts that made you a good reviewer misfire on generated code. What we replaced them with.
20 July 2026 · 9 min read
Delivery · EconomicsThe 70/30 split: what AI actually automates in a real build
Which parts of a build genuinely got cheaper, which didn't move at all, and why that distinction decides your quote.
18 July 2026 · 10 min read
Engineering · RiskIs AI-generated code safe to ship?
Not by default. Here are the six failure modes we actually see in review, and what stops each one.
16 July 2026 · 11 min read
Founders · CostWhat does it cost to build an MVP in 2026?
Real ranges, what actually drives the number, and the three ways founders overspend before they write a line of code.
14 July 2026 · 9 min read
Tell us what you’re building.
A 30-minute call with an engineer, not a salesperson. You’ll leave with a scope, a timeline and a number — or an honest reason why we aren’t the right fit.