Netdata alerts become a GREEN/YELLOW/RED verdict that every Claude Code session reads before heavy work — advisory, actionable-only, with the fix in the warning. What it watches, how agents hear it, and what a six-reviewer panel changed.
How cc_auto keeps its Claude lead on Anthropic while vendor-pinned reviewers run through OpenRouter: frontmatter model selection, request routing, credential separation, and the probes that distinguish a healthy router from a working subagent.
What each agent protocol connects, how its wire works, and when to reach for it — one diagram each.
BrowserSkill's shape in four words — CDP, profile, session, borrow — and why a daemon that cannot live in the sandbox is what makes reusing a real Chrome login possible at all.
A self-hosted coding-agent fleet, reference-style: topology, lifecycle, identities, what's baked, the numbers, the footguns, and a ranked checklist for preparing agent-friendly environments.
Most advice gives you a flat list of questions to ask your interviewer. The list isn't the problem — the round is. A peer engineer can tell you what time people left last crunch and has no idea what the promotion bar is. Matching the question to who's in the room, and the one rule for reading any answer.
The complete architecture of the autonomous coding swarm on one page — master map, loops over loops, run lifecycle, Turn-Stop handshake, git flow, admission funnel, leg lifecycle, context anatomy, harness loop, delegation, toolbox, robustness matrix, invariants.
Why agent benchmark numbers go wrong, the two paradigms of agent evaluation, a comparison of the major frameworks, and a one-adapter pattern that makes the same verifier serve both benchmarking and RL training.
Nine weeks of an agent runtime's git history reviewed commit by commit: eight eras, the five costliest incidents, the silent-failure classes, and the design lessons that survived contact with production.
An LLM-drafted design for an N×M evolve loop looked complete — then three of its load-bearing assumptions died on contact with the three repos it depends on.
Grep vs three code-graph tools on a real onboarding task: the effective ones halve the token bill — and one is worse than nothing.
Interactive and autonomous are two lanes, not one. What it takes to let a coding agent's loop close itself.
Four community optimizers, nine configurations, three studies on an autonomous coding swarm: cost is run length, not verbosity — and only one combo reliably pays for itself.