How we build, the engineering handbook in public
Brilliant Systems publishes its delivery method so clients can read it, quote it back, and check whether we kept it.
ReadInsights
Category: Delivery
Estimating, scoping and shipping. Why projects slip, what an honest timeline looks like, and how AI changes the arithmetic of building software.
What happens to the agency when the model writes the code
Software agencies built on capacity are exposed. Accountability, sector depth and delivery risk are the parts that still hold value.
ReadThe discovery phase is where agencies hide the risk
Discovery should reduce uncertainty and end in a fixed price. If it only produces a report, the risk has stayed with you.
ReadBuying outcomes instead of hours
Why paying for finished software outcomes beats paying for hours, and where fixed-price work still has real limits.
ReadRed-team week, and what it finds every single time
Before handover, we spend a week attacking our own work. It always finds issues that ordinary tests and scanners miss.
ReadThe handover is the product
Handover should leave your team able to run and change the system without us. That means code, runbooks, decision records and rehearsal.
ReadWhat a written definition of done actually contains
A useful definition of done settles delivery arguments. It states behaviours, data, interfaces, numbers, boundaries and acceptance.
ReadHow to estimate work nobody has done before
Fixed price works when estimates separate known work from real uncertainty and price discovery before commitment.
ReadThe cost curve nobody models
At AI production scale, pilot pricing breaks. Input tokens, retries, retrieval quality and human review decide whether the unit cost works.
ReadRead only for fourteen weeks, and why the plant thanked us
On manufacturing projects, we spend fourteen weeks reading before writing so the plant can trust the software before it affects the line.
Read
