⊡
No analyses yet
Click + New Analysis to run your first LOGIC Protocol query.
⊡
Nothing found
No analyses match your current filters.
Try adjusting your search term, folder, tag, mode, or date filters.
Click + New Analysis to run your first LOGIC Protocol query.
No analyses match your current filters.
Try adjusting your search term, folder, tag, mode, or date filters.
Defaults applied to every new analysis — each can be overridden per analysis.
What this does: Sync stores a secure, compressed copy of your analyses in a private GitHub "repository" — a private folder on the internet owned by you. You then run 🔄 on each device to pull that backup down and merge it with local changes. Only you can access it.
Step 1 — Create a GitHub account (skip if you have one)
1. Go to github.com and click Sign up.
2. Enter an email, a password, and a username. Verify your email when prompted.
Step 2 — Create a private repository (a private folder for your backup)
1. Click the + icon (top-right) → New repository.
2. Choose any repository name — it doesn't have to be specific; for example logic-backup.
3. Keep it Private — this is essential; "Public" would let anyone read your backup.
4. Leave everything else as-is, then click Create repository.
5. You don't need to add a README or anything else.
Step 3 — Create a "personal access token" (the app's secure key to that folder)
1. Go to github.com/settings/tokens.
2. Click Generate new token → Generate new token (fine-grained).
3. Under Token name, type anything, e.g. LOGIC Protocol.
4. Under Expiration, choose No expiration.
5. Under Repository access, choose Only select repositories and tick the repo you created in Step 2.
6. Under Permissions, find Repository permissions → Contents → set to Read and write. Nothing else is required.
7. Click Generate token at the bottom.
8. Copy the token now — it starts with github_pat_ and is shown only once. Save a copy somewhere safe (a password manager, a notes app, or a local file). If you lose it, delete and create a new one.
Step 4 — Connect the app
1. Open the ☰ menu in LOGIC Protocol and choose Settings.
2. Under GitHub Sync, fill in:
• Repo owner — your GitHub username
• Repo name — the repo name you chose in Step 2
• GitHub PAT — the token you saved in Step 3
3. Click ⬆ Sync to GitHub (no need to press Save first).
Step 5 — Done. Sync on any device uses the same owner/repo/token. Running 🔄 once per device converges everything.
Notes
• Any repo name works. You're not tied to a particular name — the one you entered in Step 4 is the one that's used.
• Using the backup to restore your PAT: after your first export, you can Export All Data and keep that file safe. On a new device, importing it restores everything, including your GitHub PAT — so you don't need to type the long token by hand again.
Troubleshooting: If sync says "Lock acquire failed" or "Not Found", the token likely lacks access to the repo — redo Step 3, making sure the repo is ticked and Contents is Read and write. If a non-expiring token is ever lost or suspected exposed, delete it in GitHub and create a new one.
The first block always shows what this device holds — repo, token length, prefix, a short fingerprint and whether it is plain ASCII. The token value is never included. Failure records add the request's body size, the response headers and the connection state. Compare the fp= line between two devices to prove they hold the same token.
No sync failures recorded on this device.
Team Share (secondary sync — separate repo)
Shares only analyses in folders toggled 🔗 under Library → 📁 Folders. Use its own repo and a fine-grained PAT scoped to that repo only — it cannot touch your personal backup.
Exclude a tag to drop analyses carrying it from every Workshop chat. Delete removes the tag from all analyses.
Soft-deleted items appear in Library → 🗑 Trash tab where you can restore or permanently delete them.
⚠️ This permanently deletes ALL local data on this device.
Removes all analyses, settings, GitHub config, folders, tags, visibility defaults, deleted-items log, theme preference, and saved login details from IndexedDB and localStorage. Export a backup first.
Loading modes…
Paragraphs per councilor based on score (0–5) and depth level. Bias pre-registration block is always 2–4 paragraphs (unaffected).
| Score | Brief | Review | Resolution |
|---|---|---|---|
| 5 | 3–5 ¶ | 6–8 ¶ + engage ≥1 alternative | 10–12 ¶ + engage ≥2 alternatives + rebut counterargument + trace full chain |
| 4 | 2–3 ¶ | 4–5 ¶ | 7–8 ¶ + engage ≥1 alternative + rebut counterargument |
| 3 | 2 ¶ | 3–4 ¶ | 5–6 ¶ + engage ≥1 alternative |
| 2 | 2 ¶ | 2–3 ¶ | 3–4 ¶ |
| 1–0 | Not invited | Not invited | Not invited |
Score 2 councilors are not invited under normal selection (threshold is 3). They only appear when pulled in by the counterweight, category coverage, or epistemological coverage rules to meet composition requirements.
Counterweight override: Counterweight councilors match the paragraph count of the councilor(s) whose lens most closely aligns with the dominant framing they are counterweighting against.
Profiles control format (chart count, table density). Atlas Depth controls how much data (section count, row depth). They work independently.
| Profile\Depth | Near Seas | High Seas | Uncharted Waters |
|---|---|---|---|
| Feature | Few tables, ~2-3 charts | More tables, ~3-4 charts | Extensive tables, ~3-5 charts |
| Archive | Few tables, ≤1 chart | More tables, ≤1 chart | Extensive tables, ≤1 chart |
| Bulletin | Few compact tables, ≤2 charts | More compact tables, ≤2 charts | Extensive compact tables, ≤2 charts |
| Source | Few tables, 0 charts | More tables, 0 charts | Extensive tables, 0 charts |
How much prior knowledge and reading load the Dreyfus version assumes. The five trade stages map to the Dreyfus stages; the method changes, not the evidence. Novice uses short sentences and common words; Grandmaster assumes field fluency.
| Trade stage | Dreyfus stage | What the text assumes |
|---|---|---|
| Novice | Novice | No prior knowledge. Short sentences, common words, every specialist term explained, a concrete case first, step-by-step signposts. |
| Apprentice | Advanced Beginner | Recognizes the topic. Key terms explained, one example per principle, light signposts. |
| Journeyman | Competent | General reader. Specialist terms on first use, examples as support, normal transitions. |
| Master | Proficient | Field-fluent. Dense prose, vocabulary assumed, finding first, examples only when load-bearing. |
| Grandmaster | Expert | Deep familiarity. Dense, technical prose; technical distinctions kept; non-obvious point first; obvious compressed. |
The Dreyfus version is always written in natural voice. Chat depth is independent of the chat Voice setting. Anvil and Workshop chats can use a different depth from the document.
Controls how many web sources the model retrieves and analyses. All tiers use full-page extracts per source (high search context), not just snippets.
The 60 / 100 / 150 figures are the initial Exa results. Each tier also funds a small follow-up pass: 2 / 3 / 4 targeted searches and 1 / 2 / 3 page fetches before the atlas, plus 1 / 2 / 3 adversarial follow-ups with 1 / 2 / 3 fetches. Follow-up retrieval is included in the credit cost and never adds a new atlas section.
| Depth | Credits | Sources | When to use |
|---|---|---|---|
| Prospecting | 1 | 60 | Quick fact-checking, familiar topics |
| Excavating | 2 | 100 | Real research — diverse current sources |
| Fracking | 3 | 150 | Maximum grounding — literature-review grade |
Each lens restructures your query differently:
| Lens | Description |
|---|---|
| 🎯 Auto | Automatically selects the best lens based on your query content |
| 📘 Academic | Frames the query for scholarly research, emphasising literature and methodology |
| 🔎 Exploratory | Broadens the query to discover unexpected connections and patterns |
| 🎯 Precise | Narrows the query for maximum specificity and focused answers |
| ⚖️ Comparative | Structures the query to compare multiple alternatives side-by-side |
| 🌀 Counterfactual | Reframes the query to explore "what if" scenarios and causal assumptions |
| Lens | Description |
|---|---|
| ✨ Clarify | States your position cleanly — strips muddle and loaded terms without changing its force |
| 🛡️ Harden | Builds the strongest defensible version of your own position, conditions folded in |
| ⚔️ Steelman Opposition | States what the opposition actually believes and wants — their real counter-argument, fairly represented |
| 💥 Stress-test | Restates your position in falsifiable form and names the single decisive test that could break it |
| Lens | Description |
|---|---|
| 🎯 Auto | Automatically selects the best data lens based on your input |
| 🗺️ Exhaustive | Maximizes data density — every measure, dimension, and boundary the topic supports |
| ⚖️ Comparative | Structures the request to compare sources, regions, or time periods side-by-side |
| ⚔️ Contested | Foregrounds divergent figures, methodology differences, and corroborated vs. single-source values |
| 📈 Temporal | Emphasizes time series, granularity, trend breaks, and periods of missing data |
| 🌐 Source Diversity | Requests data across source families — official, peer-reviewed, industry, NGO — and notes where they agree |
| 🕳️ Gaps | Foregrounds what is not known — missing periods, unmeasured populations, and where the data stops |
Select the subsections to include. Each becomes a separate sheet in the workbook.
What should this export include?
Should the Anvil summary chat be included in this export?
Running LOGIC Protocol analysis…
0m 0s elapsed
Your analyses stay saved on this device. You'll need your password to sign back in.