// Field Notes
Practical writing from real engagements — what to do when a project goes sideways, how to check code an AI wrote, what’s worth automating, and how shipping software should actually feel. No fluff, no invented numbers, nothing we haven't done.
The work everyone names from memory is the annoying work, not the expensive work. How to count the hours, then subtract the four things not worth building.
Read the post →A person is currently your integration. What “won't talk” really means, your options in order of what they cost, and the one question that decides the design.
Read the post →When an off-the-shelf automation platform is the right buy, when its per-task pricing turns into a tax, and the five questions that decide it per automation.
Read the post →Nobody publishes a price, and a padded range wouldn't help. The three shapes this work is billed in, what moves your number, and the running cost quotes miss.
Read the post →Your developer went silent. How to secure your domain, repository, hosting and data first — and where the technical work stops and a lawyer starts.
Read the post →A developer told you to rebuild it properly. Before you spend a rewrite's worth of money: which of the four layers in an AI-built app is worth keeping.
Read the post →Nobody publishes a price for making an AI-built app production ready. The three shapes this work gets billed in, and what really moves your number.
Read the post →What DevOps as a service means if you don't write code: what it covers, what it is not (an MSP, cloud hosting, outsourced developers), and when it doesn't fit.
Read the post →Nobody publishes a price, and a padded range wouldn't help. The three shapes providers bill in, what moves your number, and how to read two quotes fairly.
Read the post →The questions that separate a good DevOps provider from an expensive one — including the exit question most buyers never ask and most providers hope you won't.
Read the post →One person can deploy, and everything queues behind them. Why key-person risk builds quietly, what it costs, and why “write documentation” never fixes it.
Read the post →If you learn about outages from support emails, you don't have monitoring. The four layers worth watching, alerts that get read, and five fixes for this week.
Read the post →Why adopting Scrum after a bad month usually decays in six weeks, the three questions any working process must answer, and the smallest setup that answers them.
Read the post →What to do when a freelancer or agency abandons your project: secure your code and accounts, document everything, and avoid the expensive second mistake.
Read the post →Day one of every rescue: get it running, read the data model, mine the git history, and find the load-bearing hacks — before changing anything.
Read the post →Vibe-coded your app and now it's failing in production? The predictable ways AI-generated code breaks with real users — and what to do before you panic-rebuild.
Read the post →15 checks to run on an AI-generated codebase before real users arrive: secrets, auth, data isolation, backups, runaway queries — and how to test each one.
Read the post →Releases hurt and someone said “hire a DevOps engineer.” The honest math of the full-time hire, the real alternatives, and the cases where hiring is right.
Read the post →12 more
