AI in production. Reliability in practice.
Latest edition ·The cover story
BeginnerA café for a conversation. An old draft without disturbing today’s work. A few words change what makes a search result useful. Reranking gives them a closer look.
Read the storyFresh from the publication
A recording contains sound, not a hidden text file. Whisper turns that sound into a representation a model can use, then builds a transcript piece by piece. The task you choose helps determine what those words should be.
The assistant has your question. The release system has the answer. Follow how MCP connects the two, from a request for records to an explanation you can trace.
A four-bit model can fit where a larger copy will not. Follow one rounded weight into the calculations, memory and support cases that determine whether it helps.
The model can answer on your laptop while another part of the app sends a page elsewhere. Follow one manuscript through document reading, web research and saved copies to choose the connections you actually want.
A PHP site can serve a healthy-looking page while its worker remains exposed. Separate the runtime fixes, the worker’s reach and the investigation into what happened before repair.
An engineer asks about slow checkout. The assistant investigates, then waits for approval. Eve supplies the continuing conversation and tool machinery; your team still defines the job.
Worth understanding
Explainer
Rules, examples, prompts and reasoning changed the work we ask computers to do. The older choices still have jobs worth doing.
Guide
A promising answer is easier to review when you can find the files behind it. Follow one small repair through T3 Code’s sessions, worktrees and host machines.
Explainer
An order needs another factory, but parts, schedules and permissions live in different systems. Follow the decision through Foundry, AIP and Apollo, then see how Gotham approaches connected information.
Reliability in practice
Incidents, on-call work, and the decisions behind reliable systems.
Separate detection from response and find delays that an improving average can conceal.
Guide
A cleanup can improve backups without settling every configuration. A risk registry lets the tested improvement stand while someone takes responsibility for what remains.
Guide
Fewer primary shifts can still leave the same specialist answering the phone. Follow the interruptions, shared effort and next-day repairs before calling the workload lighter.
Put it to work
Put incident timestamps in order and examine the intervals between failure, detection, investigation, and recovery.
Build a timelineTurn a known failure into a procedure with scope, actions, escalation, and recovery checks.
Build a runbookTrack corrective actions alongside verification criteria and the risk that remains.
Track an actionStart learning
AI basics · 10 lessons available
Understand what happens between a prompt and an answer, learn how to check the result, and build a small document assistant that shows its evidence.
10 lessons ready, with practical exercises throughout.
Explore AI basicsLinux basics · 13 lessons available
Learn to operate a Linux system by building, running and troubleshooting a small service in a disposable practice environment.
13 lessons ready, with practical exercises throughout.
Explore Linux basics
