Search across Doctrine, Practice, Journal, and Community
The work while it is being done. Build logs, field reports and arguments in progress, written the day the thing happened rather than after it turned out well.
The mistakes stay in. A bug that took a day is written up as a bug that took a day; a post whose conclusion was later wrong keeps the conclusion and gets a correction under it. That is the only part of this hard to fake, and it is the reason to read it. Entries land on the Sunday Build.
31 posts
We pointed StellarView at Conduit, the most-implemented demo application on the internet, and asked for it as an operated platform rather than an app. It produced ten phases, 106 work units, and 357 questions it would not guess at.
A plan for a ten-phase platform produced 357 questions and refused to answer 281 of them. One had opened the repository, noticed docker-compose declared a build for a service whose Dockerfile did not exist, and asked who owned it. I had committed that file an hour before and left it unfinished.
Twelve hours and eighteen minutes, unattended, including two billing stops and a restart. 268 files, 25,857 lines, ten pull requests. CI green through phase eight and red on nine and ten, which we are leaving in.
Phase five wrote its own Terraform. Sixty five resources stand in a real AWS account behind a real certificate, and the application returns 503. The reason is a variable whose own description explains, in detail, why its default cannot work.
The published RealWorld suite: 13 files, 154 requests, no accommodations. It failed on the first run today because our rate limiter allows 20 anonymous writes and the suite makes 21. Then 13 of 13 in 912ms. Plus the part nobody talks about, which is who keeps a spec alive for a decade.
A field operator spent years knocking doors, understood the work precisely enough to describe it, and a factory turned that description into a running platform in weeks. Everyone is asking whether AI can build software. The harder question is who knows what to build.
The models were trained on decades of work done the hard way. Nobody has written the rules for working alongside them, because almost nobody has been here long enough to lose anything. We have. Part one of the Stella series.
A client team receives a working enterprise (platform, documentation, support desk, data lake, and a running StellarView each) before any of it is at their risk. The transfer is the last act, not the first.
Youvocate was not designed from a market study. It was built out of one person's own medical record, and the years it took to get an answer out of it. This is that record, driven through the real app.
StellarView draws a codebase as five rings that rotate and pulse. That is for observation, not decoration. Here is the full tour, with the real architecture behind each ring, from GitHub and Confluence.
