socratic-method 0.1.0__tar.gz

socratic_method-0.1.0/.github/workflows/ci.yml

@@ -0,0 +1,56 @@
+name: CI
+
+on:
+  push:
+    branches: [master, main]
+  pull_request:
+
+jobs:
+  lint:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - name: pre-commit (all files)
+        run: uvx pre-commit run --all-files --show-diff-on-failure
+
+  test:
+    runs-on: ubuntu-latest
+    strategy:
+      matrix:
+        python-version: ["3.11", "3.12", "3.13"]
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+        with:
+          python-version: ${{ matrix.python-version }}
+      - name: Sync (with dev group)
+        run: uv sync
+      - name: Tests
+        run: uv run pytest -q
+      - name: CLI smoke
+        run: |
+          uv run socratic-method --version
+          uv run socratic-method setup all --dry-run --root "$RUNNER_TEMP"
+          uv run socratic-method status --root "$RUNNER_TEMP"
+          uv run socratic-method validate evals/fixtures/tech-talk-series-20260704.md
+
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - name: Build sdist + wheel
+        run: uv build
+      - name: Check metadata
+        run: uvx twine check dist/*
+      - name: Install from wheel and smoke it
+        run: |
+          uv venv /tmp/wheel-venv
+          VIRTUAL_ENV=/tmp/wheel-venv uv pip install dist/*.whl
+          /tmp/wheel-venv/bin/socratic-method --version
+          /tmp/wheel-venv/bin/socratic-method setup all --dry-run --root "$RUNNER_TEMP"
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/

socratic_method-0.1.0/.github/workflows/release.yml

@@ -0,0 +1,75 @@
+# Tag-driven release: push a tag vX.Y.Z (matching src/socratic_method/__init__.py)
+# and this workflow builds, creates the GitHub Release with the dist files attached,
+# and publishes to PyPI via Trusted Publishing (OIDC) — no API token stored in the repo.
+#
+# One-time setup on pypi.org (repo owner):
+#   PyPI → Your account → Publishing → Add a new pending publisher:
+#     project:      socratic-method
+#     owner:        grammy-jiang
+#     repository:   socratic-method
+#     workflow:     release.yml
+#     environment:  pypi
+
+name: Release
+
+on:
+  push:
+    tags:
+      - "v*"
+  # Fallback for tags created by the "Tag release" helper workflow: tags pushed
+  # with the Actions GITHUB_TOKEN do not fire tag-push events, so dispatch this
+  # workflow manually with ref = the new tag (e.g. v0.1.0).
+  workflow_dispatch:
+
+jobs:
+  build:
+    runs-on: ubuntu-latest
+    steps:
+      - uses: actions/checkout@v4
+      - uses: astral-sh/setup-uv@v5
+      - name: Verify tag matches package version
+        run: |
+          PKG_VERSION=$(uv run python -c "import socratic_method; print(socratic_method.__version__)")
+          TAG="${GITHUB_REF_NAME}"
+          if [ "v${PKG_VERSION}" != "${TAG}" ]; then
+            echo "Tag ${TAG} does not match package version v${PKG_VERSION}" >&2
+            exit 1
+          fi
+      - name: Build sdist + wheel
+        run: uv build
+      - name: Check metadata
+        run: uvx twine check dist/*
+      - uses: actions/upload-artifact@v4
+        with:
+          name: dist
+          path: dist/
+
+  github-release:
+    needs: build
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write  # required to create the GitHub Release
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: softprops/action-gh-release@v2
+        with:
+          files: dist/*
+          generate_release_notes: true
+
+  pypi:
+    needs: build
+    runs-on: ubuntu-latest
+    environment:
+      name: pypi
+      url: https://pypi.org/p/socratic-method
+    permissions:
+      id-token: write  # required for PyPI Trusted Publishing (OIDC)
+    steps:
+      - uses: actions/download-artifact@v4
+        with:
+          name: dist
+          path: dist/
+      - uses: pypa/gh-action-pypi-publish@release/v1

socratic_method-0.1.0/.github/workflows/tag-release.yml

@@ -0,0 +1,38 @@
+# Helper for environments that cannot push tags directly (e.g. remote sandboxes
+# whose git proxy rejects refs/tags pushes). Creates the annotated release tag
+# at the dispatched ref's HEAD after checking it matches the package version.
+#
+# Note: tags pushed with the Actions GITHUB_TOKEN do not trigger tag-push
+# workflows, so after this succeeds, dispatch the Release workflow with
+# ref = the new tag.
+
+name: Tag release
+
+on:
+  workflow_dispatch:
+    inputs:
+      tag:
+        description: "Tag to create (vX.Y.Z, must match src/socratic_method/__init__.py)"
+        required: true
+        default: "v0.1.0"
+
+jobs:
+  tag:
+    runs-on: ubuntu-latest
+    permissions:
+      contents: write  # required to push the tag
+    env:
+      TAG: ${{ inputs.tag }}
+    steps:
+      - uses: actions/checkout@v4
+      - name: Create and push annotated tag
+        run: |
+          PKG_VERSION=$(sed -n 's/^__version__ = "\(.*\)"$/\1/p' src/socratic_method/__init__.py)
+          if [ "v${PKG_VERSION}" != "${TAG}" ]; then
+            echo "Tag ${TAG} does not match package version v${PKG_VERSION}" >&2
+            exit 1
+          fi
+          git config user.name "github-actions[bot]"
+          git config user.email "41898282+github-actions[bot]@users.noreply.github.com"
+          git tag -a "${TAG}" -m "socratic-method ${TAG#v} — agent skill, installer CLI, validator, evals"
+          git push origin "refs/tags/${TAG}"

socratic_method-0.1.0/.gitignore

@@ -0,0 +1,9 @@
+__pycache__/
+*.pyc
+.venv/
+dist/
+*.egg-info/
+.pytest_cache/
+uv.lock
+evals/reports/*/*/workdir/
+notes/

socratic_method-0.1.0/.pre-commit-config.yaml

@@ -0,0 +1,35 @@
+# One tool per file type, deliberately — stacked linters/formatters on the same type
+# fight each other:
+#   Python          -> ruff (lint + format in ONE tool; no black/isort/flake8 alongside)
+#   YAML            -> check-yaml (syntax only; no style formatter)
+#   GitHub workflow -> actionlint (the workflow dialect, distinct from plain YAML)
+#   JSON            -> check-json
+#   TOML            -> check-toml
+#   Markdown        -> whitespace hygiene only: a formatter would reflow
+#                      src/socratic_method/assets/SKILL.md, the shipped skill asset,
+#                      whose prose shape is deliberate.
+repos:
+  - repo: https://github.com/pre-commit/pre-commit-hooks
+    rev: v6.0.0
+    hooks:
+      - id: trailing-whitespace
+      - id: end-of-file-fixer
+      - id: mixed-line-ending
+        args: [--fix=lf]
+      - id: check-yaml
+      - id: check-json
+      - id: check-toml
+      - id: check-added-large-files
+      - id: check-merge-conflict
+
+  - repo: https://github.com/astral-sh/ruff-pre-commit
+    rev: v0.15.20
+    hooks:
+      - id: ruff-check
+        args: [--fix]
+      - id: ruff-format
+
+  - repo: https://github.com/rhysd/actionlint
+    rev: v1.7.12
+    hooks:
+      - id: actionlint

socratic_method-0.1.0/LICENSE

@@ -0,0 +1,21 @@
+MIT License
+
+Copyright (c) 2026 Grammy Jiang
+
+Permission is hereby granted, free of charge, to any person obtaining a copy
+of this software and associated documentation files (the "Software"), to deal
+in the Software without restriction, including without limitation the rights
+to use, copy, modify, merge, publish, distribute, sublicense, and/or sell
+copies of the Software, and to permit persons to whom the Software is
+furnished to do so, subject to the following conditions:
+
+The above copyright notice and this permission notice shall be included in all
+copies or substantial portions of the Software.
+
+THE SOFTWARE IS PROVIDED "AS IS", WITHOUT WARRANTY OF ANY KIND, EXPRESS OR
+IMPLIED, INCLUDING BUT NOT LIMITED TO THE WARRANTIES OF MERCHANTABILITY,
+FITNESS FOR A PARTICULAR PURPOSE AND NONINFRINGEMENT. IN NO EVENT SHALL THE
+AUTHORS OR COPYRIGHT HOLDERS BE LIABLE FOR ANY CLAIM, DAMAGES OR OTHER
+LIABILITY, WHETHER IN AN ACTION OF CONTRACT, TORT OR OTHERWISE, ARISING FROM,
+OUT OF OR IN CONNECTION WITH THE SOFTWARE OR THE USE OR OTHER DEALINGS IN THE
+SOFTWARE.

socratic_method-0.1.0/PKG-INFO

@@ -0,0 +1,170 @@
+Metadata-Version: 2.4
+Name: socratic-method
+Version: 0.1.0
+Summary: A Socratic questioning skill for coding agents (Claude Code, Codex, Copilot), with a one-command installer and an idea-brief validator.
+Project-URL: Homepage, https://github.com/grammy-jiang/socratic-method
+Project-URL: Issues, https://github.com/grammy-jiang/socratic-method/issues
+Author: Grammy Jiang
+License-Expression: MIT
+License-File: LICENSE
+Keywords: agent-skills,claude-code,codex,copilot,socratic-method
+Classifier: Development Status :: 4 - Beta
+Classifier: Environment :: Console
+Classifier: Intended Audience :: Developers
+Classifier: Programming Language :: Python :: 3.11
+Classifier: Programming Language :: Python :: 3.12
+Classifier: Programming Language :: Python :: 3.13
+Classifier: Topic :: Software Development :: Quality Assurance
+Requires-Python: >=3.11
+Requires-Dist: jsonschema>=4.0
+Requires-Dist: pyyaml>=6.0
+Description-Content-Type: text/markdown
+
+# socratic-method
+
+A **Socratic questioning skill for coding agents** — Claude Code, OpenAI Codex, and
+GitHub Copilot — packaged with a one-command installer and a deterministic artifact
+validator.
+
+Invoke the skill before real work starts (writing software, drafting a plan or document,
+making a decision) and the agent becomes a disciplined questioner: it steelmans your idea,
+interrogates it one question at a time (classic elenchus — six Socratic question types,
+counterexamples, contradiction-surfacing by quoting your own words), and ends with an
+honest verdict — **sharpened**, **aporia** (a genuinely unresolved hole, treated as a
+finding), or **refuted** (only ever out of your own mouth). The result is written down as
+a machine-validatable **idea brief** (`idea-brief-v1`) that downstream work can consume:
+open questions become a research agenda, unvalidated assumptions become a validation
+worklist.
+
+The skill was authored and hardened in
+[subagent-factory](https://github.com/grammy-jiang/subagent-factory): seven rounds of
+grounded review plus a six-cell adversarial behavioral eval (examiner vs. scripted user
+simulator, deterministic graders + independent judge), which caught and fixed real
+behavior gaps static review missed. The eval harness ships in this repo under `evals/`.
+
+## Install
+
+```bash
+pip install git+https://github.com/grammy-jiang/socratic-method
+# or, without installing:
+uvx --from git+https://github.com/grammy-jiang/socratic-method socratic-method --help
+```
+
+## Set up the skill for your agents
+
+```bash
+# auto-detect which agents are installed on this machine and configure those
+socratic-method setup
+
+# or name platforms explicitly / force all three
+socratic-method setup claude codex
+socratic-method setup all
+
+# install into your user home instead of the current project
+socratic-method setup claude --scope user
+
+# see what would happen first / check current state / remove
+socratic-method setup --dry-run
+socratic-method status
+socratic-method uninstall claude
+```
+
+With no targets, `setup` **auto-detects** installed agents and configures only those,
+printing the concrete evidence for each detection (never a bare claim):
+
+| Agent | Detection signals, in order |
+|---|---|
+| Claude Code | `claude` CLI on PATH; else `~/.claude/` config directory |
+| OpenAI Codex | `codex` CLI on PATH; else `~/.codex/` config directory |
+| GitHub Copilot | `copilot` CLI on PATH; else `gh-copilot` extension; else a `github.copilot*` VS Code extension |
+
+If nothing is detected, `setup` installs nothing and tells you how to name targets
+explicitly. `setup all` bypasses detection.
+
+`setup` is **idempotent** (content-hash comparison; an identical install reports
+"up to date"), refuses to overwrite locally modified files without `--force`, and after
+every write **reads the files back from disk before reporting success** — the skill's own
+"verify before claiming" rule, applied to its installer.
+
+### Where the skill lands
+
+| Platform | `--scope project` (default) | `--scope user` |
+|---|---|---|
+| Claude Code | `<root>/.claude/skills/socratic-method/` | `~/.claude/skills/socratic-method/` |
+| OpenAI Codex | `<root>/.agents/skills/socratic-method/` | `~/.agents/skills/socratic-method/` |
+| GitHub Copilot | `<root>/.github/skills/socratic-method/` | — |
+
+Copilot also reads a repo's `.claude/skills/`, so if the Claude project install is already
+present, the Copilot step reports "already covered" and skips (a duplicate would trigger
+twice); `--force` installs to `.github/skills/` anyway. Paths live in one data-driven
+registry (`installer.py`) — if a platform moves its skills directory, the fix is one line.
+
+## Use the skill
+
+In any agent with the skill installed:
+
+> stress-test this plan: …
+> help me think through …
+> poke holes in this idea: …
+
+Or invoke directly: `/socratic-method <idea> [--mode stress|develop] [--depth quick|standard|deep]`.
+The session ends with the brief saved to `notes/idea-briefs/<slug>-YYYYMMDD.md`.
+
+## Validate a brief
+
+```bash
+socratic-method validate notes/idea-briefs/my-idea-20260704.md
+```
+
+Checks the YAML frontmatter against the packaged `idea-brief-v1` JSON schema plus the
+cross-field rules a schema can't express (e.g. `verdict: refuted` requires the two
+colliding claims verbatim in the body; `verdict: aporia` requires open questions).
+
+## Behavioral eval harness (`evals/`)
+
+The six-cell regression matrix that hardened the skill: normal cells (planted
+contradiction → refuted; genuine unknowns → aporia; quick-depth cadence), edge cells
+(mid-session stop; disputed restatement), and an out-of-scope cell (fully specified plan →
+decline). Each cell runs a live examiner against a scripted user simulator, then grades
+the transcript with deterministic graders and an independent model judge.
+
+```bash
+python evals/run_eval.py --dry-run     # list cells, no calls
+python evals/run_eval.py --cell O1     # one cell
+python evals/run_eval.py               # full matrix — spawns ~30-60 headless `claude` calls
+```
+
+Requires the `claude` CLI and real tokens; run cells individually while iterating. A cell
+passes only when **all** deterministic graders pass AND the judge confirms the expected
+behavior with no fabrication.
+
+## Development
+
+```bash
+uv sync                      # or: pip install -e . && pip install pytest
+uv run pytest                # validator negatives, installer + detection behavior, CLI smoke
+uv run pre-commit install    # one tool per file type: ruff (py), syntax checks (yaml/json/toml), actionlint
+```
+
+CI (`.github/workflows/ci.yml`) runs the test suite on Python 3.11–3.13, builds the
+sdist/wheel, checks metadata, and smoke-tests the CLI installed from the built wheel.
+
+## Releasing to PyPI
+
+Publishing uses **PyPI Trusted Publishing** (OIDC) — no API token lives in the repo.
+One-time setup on pypi.org: *Account → Publishing → Add a new pending publisher* with
+project `socratic-method`, owner `grammy-jiang`, repository `socratic-method`, workflow
+`release.yml`, environment `pypi`. Then, to release:
+
+1. Bump `__version__` in `src/socratic_method/__init__.py`.
+2. Create a GitHub Release with tag `vX.Y.Z` (must match — the workflow verifies).
+3. `release.yml` builds, checks, and publishes; the package appears at
+   `pypi.org/project/socratic-method/`, after which `pip install socratic-method` works.
+
+Maintenance rule inherited from the skill's history: a behavioral failure is not fixed
+until something durable changes — a grader, a scenario, or a rail in `SKILL.md` — never a
+prose spot-fix alone.
+
+## License
+
+MIT

socratic_method-0.1.0/README.md

@@ -0,0 +1,148 @@
+# socratic-method
+
+A **Socratic questioning skill for coding agents** — Claude Code, OpenAI Codex, and
+GitHub Copilot — packaged with a one-command installer and a deterministic artifact
+validator.
+
+Invoke the skill before real work starts (writing software, drafting a plan or document,
+making a decision) and the agent becomes a disciplined questioner: it steelmans your idea,
+interrogates it one question at a time (classic elenchus — six Socratic question types,
+counterexamples, contradiction-surfacing by quoting your own words), and ends with an
+honest verdict — **sharpened**, **aporia** (a genuinely unresolved hole, treated as a
+finding), or **refuted** (only ever out of your own mouth). The result is written down as
+a machine-validatable **idea brief** (`idea-brief-v1`) that downstream work can consume:
+open questions become a research agenda, unvalidated assumptions become a validation
+worklist.
+
+The skill was authored and hardened in
+[subagent-factory](https://github.com/grammy-jiang/subagent-factory): seven rounds of
+grounded review plus a six-cell adversarial behavioral eval (examiner vs. scripted user
+simulator, deterministic graders + independent judge), which caught and fixed real
+behavior gaps static review missed. The eval harness ships in this repo under `evals/`.
+
+## Install
+
+```bash
+pip install git+https://github.com/grammy-jiang/socratic-method
+# or, without installing:
+uvx --from git+https://github.com/grammy-jiang/socratic-method socratic-method --help
+```
+
+## Set up the skill for your agents
+
+```bash
+# auto-detect which agents are installed on this machine and configure those
+socratic-method setup
+
+# or name platforms explicitly / force all three
+socratic-method setup claude codex
+socratic-method setup all
+
+# install into your user home instead of the current project
+socratic-method setup claude --scope user
+
+# see what would happen first / check current state / remove
+socratic-method setup --dry-run
+socratic-method status
+socratic-method uninstall claude
+```
+
+With no targets, `setup` **auto-detects** installed agents and configures only those,
+printing the concrete evidence for each detection (never a bare claim):
+
+| Agent | Detection signals, in order |
+|---|---|
+| Claude Code | `claude` CLI on PATH; else `~/.claude/` config directory |
+| OpenAI Codex | `codex` CLI on PATH; else `~/.codex/` config directory |
+| GitHub Copilot | `copilot` CLI on PATH; else `gh-copilot` extension; else a `github.copilot*` VS Code extension |
+
+If nothing is detected, `setup` installs nothing and tells you how to name targets
+explicitly. `setup all` bypasses detection.
+
+`setup` is **idempotent** (content-hash comparison; an identical install reports
+"up to date"), refuses to overwrite locally modified files without `--force`, and after
+every write **reads the files back from disk before reporting success** — the skill's own
+"verify before claiming" rule, applied to its installer.
+
+### Where the skill lands
+
+| Platform | `--scope project` (default) | `--scope user` |
+|---|---|---|
+| Claude Code | `<root>/.claude/skills/socratic-method/` | `~/.claude/skills/socratic-method/` |
+| OpenAI Codex | `<root>/.agents/skills/socratic-method/` | `~/.agents/skills/socratic-method/` |
+| GitHub Copilot | `<root>/.github/skills/socratic-method/` | — |
+
+Copilot also reads a repo's `.claude/skills/`, so if the Claude project install is already
+present, the Copilot step reports "already covered" and skips (a duplicate would trigger
+twice); `--force` installs to `.github/skills/` anyway. Paths live in one data-driven
+registry (`installer.py`) — if a platform moves its skills directory, the fix is one line.
+
+## Use the skill
+
+In any agent with the skill installed:
+
+> stress-test this plan: …
+> help me think through …
+> poke holes in this idea: …
+
+Or invoke directly: `/socratic-method <idea> [--mode stress|develop] [--depth quick|standard|deep]`.
+The session ends with the brief saved to `notes/idea-briefs/<slug>-YYYYMMDD.md`.
+
+## Validate a brief
+
+```bash
+socratic-method validate notes/idea-briefs/my-idea-20260704.md
+```
+
+Checks the YAML frontmatter against the packaged `idea-brief-v1` JSON schema plus the
+cross-field rules a schema can't express (e.g. `verdict: refuted` requires the two
+colliding claims verbatim in the body; `verdict: aporia` requires open questions).
+
+## Behavioral eval harness (`evals/`)
+
+The six-cell regression matrix that hardened the skill: normal cells (planted
+contradiction → refuted; genuine unknowns → aporia; quick-depth cadence), edge cells
+(mid-session stop; disputed restatement), and an out-of-scope cell (fully specified plan →
+decline). Each cell runs a live examiner against a scripted user simulator, then grades
+the transcript with deterministic graders and an independent model judge.
+
+```bash
+python evals/run_eval.py --dry-run     # list cells, no calls
+python evals/run_eval.py --cell O1     # one cell
+python evals/run_eval.py               # full matrix — spawns ~30-60 headless `claude` calls
+```
+
+Requires the `claude` CLI and real tokens; run cells individually while iterating. A cell
+passes only when **all** deterministic graders pass AND the judge confirms the expected
+behavior with no fabrication.
+
+## Development
+
+```bash
+uv sync                      # or: pip install -e . && pip install pytest
+uv run pytest                # validator negatives, installer + detection behavior, CLI smoke
+uv run pre-commit install    # one tool per file type: ruff (py), syntax checks (yaml/json/toml), actionlint
+```
+
+CI (`.github/workflows/ci.yml`) runs the test suite on Python 3.11–3.13, builds the
+sdist/wheel, checks metadata, and smoke-tests the CLI installed from the built wheel.
+
+## Releasing to PyPI
+
+Publishing uses **PyPI Trusted Publishing** (OIDC) — no API token lives in the repo.
+One-time setup on pypi.org: *Account → Publishing → Add a new pending publisher* with
+project `socratic-method`, owner `grammy-jiang`, repository `socratic-method`, workflow
+`release.yml`, environment `pypi`. Then, to release:
+
+1. Bump `__version__` in `src/socratic_method/__init__.py`.
+2. Create a GitHub Release with tag `vX.Y.Z` (must match — the workflow verifies).
+3. `release.yml` builds, checks, and publishes; the package appears at
+   `pypi.org/project/socratic-method/`, after which `pip install socratic-method` works.
+
+Maintenance rule inherited from the skill's history: a behavioral failure is not fixed
+until something durable changes — a grader, a scenario, or a rail in `SKILL.md` — never a
+prose spot-fix alone.
+
+## License
+
+MIT

socratic_method-0.1.0/evals/fixtures/tech-talk-series-20260704.md

@@ -0,0 +1,55 @@
+---
+schema: idea-brief-v1
+idea: tech-talk-series
+date: 2026-07-04
+mode: stress
+depth: standard
+verdict: sharpened
+thesis_final: "A monthly 30-minute internal tech talk, seeded with 3 pre-committed speakers,
+  to improve cross-team visibility — piloted for a quarter before any cadence increase."
+questions_asked: 9
+types_used: [clarification, assumptions, evidence, implications]
+assumptions:
+  - text: "Cross-team visibility is a widely felt problem, not just the proposer's"
+    status: unvalidated
+  - text: "Enough engineers will volunteer to speak at monthly cadence"
+    status: risky
+  - text: "30 minutes is short enough that attendance survives busy weeks"
+    status: unvalidated
+open_questions:
+  - "Do at least 5 other people report being surprised by another team's work in the last quarter?"
+  - "Who are speakers 1-3, by name, before anything is announced?"
+constraints:
+  - "No budget; must run on existing meeting infrastructure"
+  - "Cannot be mandatory (user: 'the moment it's mandatory it's dead')"
+next_step: "Ask the last 5 people who joined what they wish they'd known — validates the
+  visibility assumption before committing anyone's Friday."
+---
+
+# Idea brief: internal tech-talk series
+
+## What changed under questioning
+Initial: weekly talks to fix visibility and grow juniors, weekly cadence essential.
+Final: monthly pilot with pre-committed speakers; the junior-growth goal was dropped as a
+separate untested claim riding along with the visibility one.
+
+## Scope
+For: engineers across teams at the user's company. Out of scope: external audience,
+recorded-content library, mandatory attendance.
+
+## Assumptions surfaced
+The visibility problem rests on a sample of one (the user's own migration surprise). Speaker
+supply was the load-bearing weakness: the user's own words — "I'd be begging people" —
+forced the cadence change.
+
+## Contradictions & how resolved
+"Weekly, otherwise it loses momentum" vs "most engineers here hate presenting" — resolved by
+the user: cadence yielded ("monthly with strong talks beats weekly with filler").
+
+## Open questions (aporia)
+Whether the visibility problem is real beyond the user, and who the first three named
+speakers are. Both must be answered before announcing anything.
+
+## Suggested next step
+Interview the last 5 joiners about what they wish they'd known. Cheap, fast, and directly
+tests the founding assumption.