Enginery
Live Is it working for real people?
Cycle 0 was called at day 10 and produced one finding worth the whole exercise. Sorted by what each key result actually required, the split was perfect: five of five that needed only that the code get written scored full marks, and six of six that needed a stranger to do something scored close to zero.
So every key result on this card is a reach key result. None of them can be closed by writing more software, which is deliberate, and uncomfortable, and the point.
The strategy underneath this makes one sharp prediction: that positioning in front of demand that already exists will beat messaging an audience I assembled myself, on arrivals per hour spent, by a wide margin. Both halves are measurable inside a fortnight. E1.1 is one half and E2.2 is the other, which is the only reason a channel I have otherwise retired still appears on this board.
E1 Find out whether one asset can intercept demand
The whole operating playbook rests on an unmeasured assumption: that one asset can intercept meaningfully more than 29 downloads a month, which is the best rate anything here has ever produced. The App Store optimisation pass on Pet Med Reminder measures exactly that, and nothing else scales until the number exists. The registry listing is the cheapest untested idea in the portfolio: a one-time act that produces a permanently indexed page, spending a machine's hours forever rather than mine.
| # | Key result | Baseline | Target | Where it is |
|---|---|---|---|---|
| E1.1 | Impressions reach 320 in the ten days after 1.9.0 | 614 per 30 days, roughly 205 per 10 days | 320 | Not yet read: Not started |
| E1.2 | Written verdict on the App Store channel by 13 Sept | No channel has ever been run deliberately to a threshold | Written | Not yet read: Not started |
| E1.3 | One ecosystem registry listing published by 7 Sept | Zero registries ever tried, with sixteen tools sitting unpublished | 1 | Not yet read: Not started |
E2 Attribute the inflow
Two strangers arrived at the Italian word game in late August. One of them came back the next day and finished a test he had abandoned three times, which is the only retention evidence this portfolio has ever produced. Nobody knows how either of them got there. It was not the share links and it was not Reddit. The second and third key results here exist because the strategy this cycle runs on makes one sharp prediction, and both halves of it have to be measurable for the prediction to be worth anything.
| # | Key result | Baseline | Target | Where it is |
|---|---|---|---|---|
| E2.1 | Arrival source recorded for every external session | Unattributed | 100% | Not yet read: Not started |
| E2.2 | Arrivals from the August group message read by 5 Sept | The message went out on 29 Aug, tagged, and has never been read | Read | Not yet read: Not started |
| E2.3 | Analytics live on this site by 2 Sept | None. This site has never measured anything | Live | Not yet read: Not started |
E3 Ship into the AI-native engineering lane
Six candidate audiences were checked against the question of whether their job is already served. Five came back closed, several by four or more shipped competitors. One came back open: attribution of AI-authored code for engineering organisations, where every existing tool is enterprise and sold by a salesperson. It carried no key results at all last cycle.
| # | Key result | Baseline | Target | Where it is |
|---|---|---|---|---|
| E3.1 | Attribution tool runnable by a stranger by 10 Sept | Does not exist | Shipped, public URL | Not yet read: Not started |
| E3.2 | 3 engineering leaders run it on their own repository | 0 | 3 | Not yet read: Not started |
Get updates
Occasional notes on what shipped and what the numbers said. No schedule, no marketing.