From fbd28f560684eb3c6e08e0e573971a56f25d2d06 Mon Sep 17 00:00:00 2001
From: Claude <noreply@anthropic.com>
Date: Tue, 18 Aug 2026 15:47:56 +0000
Subject: [PATCH] prompt: route inference-method-cards-seed (#140) ->
 library-dev (cloud)

Co-Authored-By: Claude <noreply@anthropic.com>
Claude-Session: https://claude.ai/code/session_018dV9h5h76hZ6trGZKWJFHe
---
 active.md                             | 12 +++++++++
 active/inference_method_cards_seed.md | 39 +++++++++++++++++++++++++++
 2 files changed, 51 insertions(+)
 create mode 100644 active/inference_method_cards_seed.md

diff --git a/active.md b/active.md
index fd9a4176..6c64378a 100644
--- a/active.md
+++ b/active.md
@@ -1,5 +1,17 @@
 # Active Tasks
 
+## inference-method-cards-seed
+- issue: https://github.com/PyAutoLabs/autolens_profiling/issues/140
+- session: Claude Code cloud/mobile (no local worktree)
+- status: library-dev
+- worktree: none — cloud session works on the GitHub clone directly
+- repos:
+  - autolens_profiling: feature/inference-method-cards-seed
+- prompt: active/inference_method_cards_seed.md
+- note: docs-only — seed methods/{nautilus,nss,multi_start_prodigy,multi_start_adam}.md
+  from the existing benchmark record (PROGRAMME §6 template) + tick Phase 0(e) in
+  the PROGRAMME phase table (recorded follow-up from searches-readme-dashboard).
+
 ## stored-sample-reconstruction-guard
 - issue: https://github.com/PyAutoLabs/PyAutoFit/issues/1486
 - status: library-dev — WORKSPACE HALF SHIPPED; the PyAutoFit hardening (#1486) is what remains
diff --git a/active/inference_method_cards_seed.md b/active/inference_method_cards_seed.md
new file mode 100644
index 00000000..88de6eee
--- /dev/null
+++ b/active/inference_method_cards_seed.md
@@ -0,0 +1,39 @@
+# Seed the inference method evidence cards from the existing benchmark record
+
+Type: docs
+Target: autolens_profiling
+Repos:
+- @autolens_profiling
+Difficulty: small
+Autonomy: supervised
+Priority: medium
+
+The inference programme's knowledge structure
+(`results/notes/inference/PROGRAMME.md` §6) calls for living method cards at
+`results/notes/inference/methods/<method>.md`, from which the eventual
+`autolens_workspace/scripts/guides/searches/` decision tree is generated.
+None exist yet, but the evidence to seed them already does: the searches
+benchmark artifacts (now surfaced by the fixed dashboard, #139), the
+programme's state-of-play tables, the clipper-campaign notes, and the
+point-source defaults campaign.
+
+Task: seed four cards using the PROGRAMME §6 template (IDENTITY / EVIDENCE /
+STRENGTHS & WEAKNESSES / CONFIGURATION / TERMINATION / HAZARDS / PERFORMANCE /
+RECOMMENDED / REFERENCES), every claim citing a result JSON, notes file, or
+PROGRAMME section — no new claims:
+
+- `methods/nautilus.md` — the CPU reference engine; truth bars on
+  imaging cells; the cluster/point-source answer today.
+- `methods/nss.md` — BlackJAX nested slice sampling; fork-era benchmarks,
+  the +7-13 nat logZ bias hypothesis, mainline 1.6 status; Gate A pending.
+- `methods/multi_start_prodigy.md` — basin-hit story (seed 0/1 split),
+  clipper/step-scaling arc, auto-convergence limits; Phase 3 pending.
+- `methods/multi_start_adam.md` — p_hit ≈ 0.18/start, 128-start A100
+  timing, the GIGA-Lens comparison.
+
+Also tick Phase 0(e) as complete (PR#139) in PROGRAMME.md's phase-state
+table (recorded follow-up from the searches-readme-dashboard completion).
+
+Confidence tags per the programme: anecdote / seeded / gated. Documentation
+only — no runs, no scripts, no source changes; ruff untouched; do not edit
+inside auto-table sentinels.