Issues

Everything in the way, in order

One list, ranked hardest first. Not a backlog and not a roadmap: a backlog is a place things go to be forgotten, and this is the list I read every morning before deciding what to do. The row at the top is the next thing.

17 open · read 03 Sept 2026

1

Three key results sit behind one App Store queue; 1.9.0 is not staged

The current release has been in the review queue past its usual budget. Stage the next one and submit it the hour this clears.

Blocks: E1, P1 Open 3d
2

An unfinished administrative item on the account the apps publish under closes the automated store-figures read across every app, and it carries a deadline this month

Confirmed today that this is what closes the automated path under row 14, and that it is account-wide rather than one app. Finish the outstanding account paperwork.

Blocks: E1 (via row 14), plus I1 and P1 Open 5d
3

Ecosystem registry listing: zero attempts ever, the cheapest untested hypothesis in the portfolio

Fourth run at zero attempts. Pick a registry and publish one listing.

Blocks: E2 (due Sept 7, 4 days) Open 3d
6

Show Notes has no per-account cost ceiling. It burned $150 once and that is why the product parked

Untouched 5 days. Ship the cap and a test that trips it.

Blocks: S1 (due Sept 7, 4 days) Open 4d
7

The analytics setup for one site was never registered and carries no internal filter, so two automated scanner visits are being counted as real people

Register the missing project, set the filter, remove the false rows.

Blocks: clean enginery.co numbers Open 3d
8

Headless agents are minting fake players in one product, and the same signature now shows on a second

Widened today: all of the second product's recent arrivals carry the signature, and a parked product took one more anonymous row. Add the automated-traffic detector to the daily read of new arrivals.

Blocks: data quality under every Italian number, and now a second product's evidence Open 10d
9

Show Notes signups are rare, and the ones that arrive are not getting through the first run

Identify the specific defect and file it.

Blocks: none this cycle (product quality) Open 10d
10

enginery.co pre-announcement pass never ran (mail test, Slack unfurl, proofread)

Run the pass.

Blocks: publishing hygiene Open 11d
11

Public scorecard page still renders the dead Vine Trails board

Take the dead board out of the public rendering.

Blocks: publishing hygiene at the Sept 13 grade Open 11d
12

The developer account every iOS listing publishes under is still registered to a prior legal entity rather than to Enginery

Investigate the transfer path.

Blocks: run-the-business (slow burn; touches every iOS listing) Open 11d
13

enginery.co needs a simpler shape: sync the hand-written site copy back to the vault, then rebuild the sections around cycles, the board, the KPIs, the issues and a journal

Part one, the vault sync, comes first.

Blocks: run-the-business (the publishing surface); deliberately not now, the local system proves out first Open 3d
14

Paroletta's iOS store figures have never been read into a daily pulse, so the key result that names them has no series behind it

The automated path to this number is closed, so it has to be read by hand. 1.6 is live, so the series finally exists to read.

Blocks: E1, I1 Open 2d
15

Pet Med Reminder requires iOS 26.5, which nobody chose; an OS floor caps how many impressions can convert

Confirm the deployment target and lower it if it was not deliberate

Blocks: interpretation of P1 Open 7d
16

Paroletta Android has no public listing; the app is still in closed testing and I2 has nothing to grade until it reaches production

Flipped red: leaving closed testing and passing a production review has to fit inside 10 days and no step has started.

Blocks: I2 (due Sept 13, 10 days) Open 2d
17

The journal has nowhere to be read. The writing end is settled and the entries exist; the site has no journal surface

Build the surface and honour the publish gate, one entry at a time.

Blocks: run-the-business (the publishing surface) Open 2d
18

A portfolio-wide arrivals metric was computed over the wrong window, and a second block of numbers was named for units it did not hold

Narrowed. The window fault was fixed in yesterday's run and re-verified independently on every project today, so the numbers are sound. What is left is the written description of the query in two places, plus renaming the mislabelled metric.

Blocks: the intercept-rate KPI, and the credibility of every download-shaped row (E1, I1, I2, P1) Open 1d
19

A work item this list retired as done still reads open and urgent in the tracker, so it tops every urgent sweep

Close it or say what is left in it.

Blocks: nothing; tracker hygiene Open 0d

Off the list

Rows leave the list two ways. Some are done. Others stop being true, because the thing they were written against was retired underneath them, and those are worth keeping visible: a list you can only leave by succeeding is a list that has been edited.

Retired

2026-09-01: three rows retired as part of the cycle rewrite, not closed as done. A kill-verdict publish the revised set no longer grades, whose underlying work continues outside the cycle. A drop-off read that blocked a verdict row which was itself cut, because it graded engagement of people already in the door rather than discovery. And the finding that triggered the rewrite: ten-day windows cannot close before the grade date, which flat cycle-window targets now remove structurally.
2026-09-01, rank 1: done. The Cycle 1 set carries the reframed rows, and the site and the doc agree. (Noted 2026-09-03: the tracker still reads it as open, which is why row 19 exists.)
2026-09-03, rank 5: Paroletta 1.6 staged and unsubmitted, retired. 1.6 cleared review overnight, so the listing rewrite is live on the store and the row's grounds are gone. I1 moved from red to amber on it. The other half of that row, reading the download series it unlocks, lives on further down the list.
2026-09-03, rank 4: the outstanding job checks for the attribution lane, retired as done. All six ran, five of the six candidate jobs failed, and the key result is met. What the checks produced is not a tool: the surviving job clears the two filters this key result names and fails the channel and whose-hours halves of the portfolio gate, so the named next step is one piece of writing against a bar of 29 arrivals in 30 days, archived rather than iterated on if it misses.

Closed by the Aug 31 seeding review

One item cancelled as stale twice over: the product it named had been re-staged as personal, and the date it worked toward was no longer live.
One item closed as done: the Cycle 1 set adopted on Aug 30 is the set it asked for.
One newsletter property retired from the measurement map and the daily read after 19 days at zero subscribers. The mailing setup and the inbound address stay live.
A malformed field written into 20 daily records since Aug 12, repaired in place the same session it was found. It never boarded the list.

Flagged, not filed

One item was written against a grade date that has since been retired, and against two key results from a closed plan. Its substance still holds, but no current row grades it and the register has it as run-the-business. It does not board the list until that is decided.

Why the numbers skip

Rank numbers are never reused. When a row is retired its number goes with it, so the gaps in the sequence are real and a reference written a week ago still points at the same row.

The list is maintained daily, but only I re-rank it. A new issue enters at the bottom of its band and stays there until I move it, which stops the ranking from being quietly rewritten by whatever felt urgent that morning.