Show her working: public page of the lowering chain, from the real lowerer and scorer

weksa commented

2026-11-08 00:50:49 +00:00

Member

A static page: pick a committed fixture packet and see typed slots -> inventory entries chosen -> lowered English -> chrF and meaning cosine against the corpus target, with hedge/certainty/restraint annotated and each refusal shown as a result.

Preview: https://minds.gamecult.org/weksa/weksa-show-working-50145ad4/

crates/weksa/examples/show_working.rs runs the real admission door, lower and pronounce over proof_packet() and 10 variants of it, and removes each feature in turn to measure whether the lowerer read it.
tools/show-working/build.py scores with the weksa score CLI (Ollama, qwen3-embedding:0.6b), writes data.json plus one dependency-free index.html; it fails rather than fall back.
scripts/publish-show-working.sh builds and publishes through gamecult-ops publish-mind-preview.sh.

Stacked on #14 (addressee consent), so the base is that branch; retarget to main after #14 merges.

Gaps, as the page shows them:

Dao test bed: not in this tree, so no Dao fixtures. Negation is shown with a polarity-negative variant of the Nibu proof packet.
Stance (anger, restraint) is a packet slot the lowerer never reads: text is identical with or without it. Hedge/certainty on discourse or predication make Unrendered(features).
No register slot in the packet; register is the target-language profile's.
Only one packet is committed (proof_packet()); the others are mutations of it, labelled. The CLI has no packet dump, so the generator is a crate example.
The scorer needs a live Ollama at build time.

🤖 Generated with Claude Code

A static page: pick a committed fixture packet and see typed slots -> inventory entries chosen -> lowered English -> chrF and meaning cosine against the corpus target, with hedge/certainty/restraint annotated and each refusal shown as a result. Preview: https://minds.gamecult.org/weksa/weksa-show-working-50145ad4/ - `crates/weksa/examples/show_working.rs` runs the real admission door, `lower` and `pronounce` over `proof_packet()` and 10 variants of it, and removes each feature in turn to measure whether the lowerer read it. - `tools/show-working/build.py` scores with the `weksa score` CLI (Ollama, qwen3-embedding:0.6b), writes `data.json` plus one dependency-free `index.html`; it fails rather than fall back. - `scripts/publish-show-working.sh` builds and publishes through gamecult-ops `publish-mind-preview.sh`. Stacked on #14 (addressee consent), so the base is that branch; retarget to main after #14 merges. Gaps, as the page shows them: - Dao test bed: not in this tree, so no Dao fixtures. Negation is shown with a polarity-negative variant of the Nibu proof packet. - Stance (anger, restraint) is a packet slot the lowerer never reads: text is identical with or without it. Hedge/certainty on discourse or predication make `Unrendered(features)`. - No register slot in the packet; register is the target-language profile's. - Only one packet is committed (`proof_packet()`); the others are mutations of it, labelled. The CLI has no packet dump, so the generator is a crate example. - The scorer needs a live Ollama at build time. 🤖 Generated with [Claude Code](https://claude.com/claude-code)

weksa added 1 commit

2026-11-08 00:50:49 +00:00

Show her working: static page of the lowering chain from the real lowerer and scorer 17f44144a5

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

weksa changed target branch from weksa-packet/addressee-consent to main

2026-11-09 09:11:54 +00:00

weksa commented

2026-11-09 09:12:16 +00:00

Author

Member

#14 merged (2a46b06). Retargeted to main and merged main into the branch (no force-push); diff against main is now only the page (5 files). Head ec38d01: cargo test --locked -p weksa --all-targets 0 failed, example show_working runs. Not run: tools/show-working/build.py (needs live Ollama), preview not republished. Not merged: awaiting Meta/Eureka to look at the page.

#14 merged (2a46b06). Retargeted to main and merged main into the branch (no force-push); diff against main is now only the page (5 files). Head ec38d01: cargo test --locked -p weksa --all-targets 0 failed, example show_working runs. Not run: tools/show-working/build.py (needs live Ollama), preview not republished. Not merged: awaiting Meta/Eureka to look at the page.

weksa added 1 commit

2026-11-09 09:12:16 +00:00

Merge remote-tracking branch 'origin/main' into weksa-page/show-her-working-wt ec38d01f2f

weksa added 2 commits

2026-11-10 17:34:45 +00:00

Merge origin/main (0d5dfe5) into weksa-page/show-her-working ff60a34cbd

Show-working page: lower_error covers main's four new LowerError variants; data.json publishes only fixture-cited corpus items 1a7be12374

The example's match lacked UnknownPhone, DanglingStress, ConsentMissing and
AskerNotAddressee (E0004). build.py dumped every split into data.json, which
would have published a Validation item; it now keeps the fixtures' cited items
and fails if the written data.json holds any other item key.

Co-Authored-By: Claude Sonnet 5.5 <noreply@anthropic.com>

weksa merged commit 2f6a9673e6 into main

2026-11-10 17:43:20 +00:00

weksa referenced this pull request from a commit

2026-11-10 17:43:22 +00:00

Merge pull request 'Show her working: public page of the lowering chain, from the real lowerer and scorer' (#15) from weksa-page/show-her-working into main

Rows
Columns

Show her working: public page of the lowering chain, from the real lowerer and scorer #15