is this the thing that best serves the gate right now, or the most interesting thing to build?
and answers it with an assertion: /bygge exists because the maintainer had a video he actually wanted analyzed, "which is genuine pull rather than speculation".
That is one person wanting one thing once. ADR-016's Stage-0 gate measures something narrower and harder: does someone return unprompted over 3–4 weeks. Wanting a thing built and returning to it are different behaviours, and the epic's own gate only credits the second.
Those two spikes measure feasibility, and feasibility is not in doubt: infra#274 proved the transcription hop end to end, and scripts/spike-media/ (committed) proves the analysis and the page. What they would consume is three afternoons and roughly the same calendar the gate measures over — while producing no evidence about demand at all.
They also make the honest answer more expensive to give. After three afternoons and a working pipeline, "defer this" costs sunk work to say. Right now it costs nothing.
Method
Send the existing tapir.d-ma.be/bygge URL to the people whose walkthrough it is — the colleague in the recording, and anyone else who was on site or is making decisions on that project. Say what it is, ask nothing, request no feedback. A request for feedback manufactures the visit this issue is trying to observe.
Then read the access log.
Distinct opens per recipient over 14 days. The signal is a second open by the same person, on a later day, unprompted. The first open is politeness; the second is use.
Non-goals
Analytics, tracking pixels, per-user tokens. The nginx access log is enough and already exists.
Asking people what they think of it. Stated intent is not the measurement.
Building anything. No script, no code, no repo change.
Enlarging the sample. Two or three real recipients on a real project beats a survey.
Acceptance
caveman: me find out if anyone come back to look again, before me build more
URL sent to ≥2 real recipients; date and recipients recorded here
Access-log count of distinct opens per recipient over 14 days, posted as a comment
Any recipient opens it a second time on a later day → pull is real. Unblock #29/#30, write #28's ADR.
Zero second opens → record in #28 that the honest answer to its own sequencing question is "the most interesting thing to build", and defer the epic behind whatever the gate says users actually do.
Either way the answer is written into #28, not just into this issue
Refs
#28 (parent — the sequencing question this answers), ADR-016 (Stage-0 gate), ADR-020 (search/facets deferred as premature against the same gate)
#29, #30 (blocked on this), #31 (not blocked — it is a chore, not a spike)
scripts/spike-media/ — the toolchain, recovered and committed, so nothing is lost by waiting
One afternoon, zero engineering. Blocks #29 and #30. Parent: #28.
## The question this actually answers
#28 asks itself, in writing:
> **is this the thing that best serves the gate right now, or the most interesting thing to build?**
and answers it with an assertion: `/bygge` exists because the maintainer had a video he actually wanted analyzed, *"which is genuine pull rather than speculation"*.
That is one person wanting one thing once. ADR-016's Stage-0 gate measures something narrower and harder: **does someone return unprompted over 3–4 weeks.** Wanting a thing built and returning to it are different behaviours, and the epic's own gate only credits the second.
So test the claim instead of restating it.
## Why this must run before #29 and #30
Those two spikes measure **feasibility**, and feasibility is not in doubt: infra#274 proved the transcription hop end to end, and `scripts/spike-media/` (committed) proves the analysis and the page. What they would consume is three afternoons and roughly the same calendar the gate measures over — while producing no evidence about demand at all.
They also make the honest answer more expensive to give. After three afternoons and a working pipeline, "defer this" costs sunk work to say. Right now it costs nothing.
## Method
Send the existing `tapir.d-ma.be/bygge` URL to the people whose walkthrough it is — the colleague in the recording, and anyone else who was on site or is making decisions on that project. Say what it is, ask nothing, request no feedback. A request for feedback manufactures the visit this issue is trying to observe.
Then read the access log.
Distinct opens per recipient over 14 days. **The signal is a second open by the same person**, on a later day, unprompted. The first open is politeness; the second is use.
## Non-goals
- Analytics, tracking pixels, per-user tokens. The nginx access log is enough and already exists.
- Asking people what they think of it. Stated intent is not the measurement.
- Building anything. No script, no code, no repo change.
- Enlarging the sample. Two or three real recipients on a real project beats a survey.
## Acceptance
- [ ] `caveman: me find out if anyone come back to look again, before me build more`
- [ ] URL sent to ≥2 real recipients; date and recipients recorded here
- [ ] Access-log count of distinct opens per recipient over 14 days, posted as a comment
- [ ] **Any recipient opens it a second time on a later day** → pull is real. Unblock #29/#30, write #28's ADR.
- [ ] **Zero second opens** → record in #28 that the honest answer to its own sequencing question is "the most interesting thing to build", and defer the epic behind whatever the gate says users actually do.
- [ ] Either way the answer is written into #28, not just into this issue
## Refs
- #28 (parent — the sequencing question this answers), ADR-016 (Stage-0 gate), ADR-020 (search/facets deferred as premature against the same gate)
- #29, #30 (blocked on this), #31 (not blocked — it is a chore, not a spike)
- `scripts/spike-media/` — the toolchain, recovered and committed, so nothing is lost by waiting
Blocking a user prevents them from interacting with repositories, such as opening or commenting on pull requests or issues. Learn more about blocking a user.
One afternoon, zero engineering. Blocks #29 and #30. Parent: #28.
The question this actually answers
#28 asks itself, in writing:
and answers it with an assertion:
/byggeexists because the maintainer had a video he actually wanted analyzed, "which is genuine pull rather than speculation".That is one person wanting one thing once. ADR-016's Stage-0 gate measures something narrower and harder: does someone return unprompted over 3–4 weeks. Wanting a thing built and returning to it are different behaviours, and the epic's own gate only credits the second.
So test the claim instead of restating it.
Why this must run before #29 and #30
Those two spikes measure feasibility, and feasibility is not in doubt: infra#274 proved the transcription hop end to end, and
scripts/spike-media/(committed) proves the analysis and the page. What they would consume is three afternoons and roughly the same calendar the gate measures over — while producing no evidence about demand at all.They also make the honest answer more expensive to give. After three afternoons and a working pipeline, "defer this" costs sunk work to say. Right now it costs nothing.
Method
Send the existing
tapir.d-ma.be/byggeURL to the people whose walkthrough it is — the colleague in the recording, and anyone else who was on site or is making decisions on that project. Say what it is, ask nothing, request no feedback. A request for feedback manufactures the visit this issue is trying to observe.Then read the access log.
Distinct opens per recipient over 14 days. The signal is a second open by the same person, on a later day, unprompted. The first open is politeness; the second is use.
Non-goals
Acceptance
caveman: me find out if anyone come back to look again, before me build moreRefs
scripts/spike-media/— the toolchain, recovered and committed, so nothing is lost by waiting