Author
Dani Reyes writes DevMoment field notes on AI dev workflow, tested on real work rather than demos.
The Batch API's default prompt cache expires in 5 minutes while the batch itself runs for up to an hour, so the two cost levers quietly cancel each other out. Six documented constraints, and the single assumption underneath all of them. Field log, September 2026.
Claude Code auto compact fires on a per-model token boundary, not a percentage. What it costs is set by prompt cache warmth rather than context size, and a path-scoped rule does not come back afterwards until Claude reads a matching file again.
Claude Code's OpenTelemetry export redacts prompts and responses by default. It does not redact who you are. Four identity attributes ship on every datapoint with no switch to turn them off, and one environment variable quietly controls two streams.
I went looking for which model was answering me and found three separate precedence orders, fourteen documented inputs, and two environment variables that resolve in opposite directions. Read from Anthropic's own docs on 9 September 2026, with the version numbers.
Installing Claude Code GitHub Actions is two decisions, and the app install is the one that sticks. The Claude GitHub App grants 11 repository permissions, the action uses 3, you cannot accept a subset, and the set can grow later. A counted field log.
The Cost column in ccusage is a counterfactual: what your tokens would have cost at pay-per-use list prices, computed locally. On a Claude subscription it is not your bill, and ccusage's own introduction and Anthropic's own docs both say so. A field log on what the number measures, where the caveat is written, and where it is missing.
The MCP 2026-07-28 revision marks the Sampling feature deprecated under SEP-2577, eight months after the revision that expanded it. A field log on what the deprecation registry actually says, why elicitation survived and sampling did not, and what to check before rewriting anything.
Elicitation has been in MCP since June 2025, but the current 2026-07-28 revision changed how it is delivered and calls the old pattern a breaking change. A field log on the three-revision diff, why elicitationId became requestState, and why the statefulness advice inverted.
MCP Apps went stable as the first official MCP extension in January 2026. A pre-build field log on the identifier mismatch, the bilateral opt-in that fails silently, the deny-by-default CSP, and which hosts actually render one in August 2026.
Every explainer quotes a 200,000-token ceiling. Anthropic's own docs say the auto-compact window is a setting you choose, from 100K to 1M, and that Sonnet 5 compacts at about 967K. The popular fix is to disable auto-compact. There is a flag instead.
Most of page one still compares Devin the autonomous cloud agent against Claude Code the terminal agent. I opened both 2026 pricing pages and the billing docs, and two of the claims everyone repeats are out of date.
I built the same CSV-export feature with GitHub Copilot and Claude Code from the same commit. The code came out close. The pricing shape is what actually decides it.
