Toolbar

Send To →

Open provider tabs: ChatGPT, Claude, Gemini, Grok. Click a button to launch; copies aren’t auto-sent (privacy-friendly).

Open target after copy

When checked, any “Copy …” action opens your selected provider automatically.

ADHD view

Compact Cognitive Map output for faster scanning in SCM-11 prompts. Toggle on/off anytime.

Transcript

Paste or Drop .txt

Box accepts raw transcripts. Timecodes like [00:12:31] are detected and used for splits.

Auto Plan

SP-22 planner estimates tokens and proposes 1–6 segments (hard cap 6 for simplicity & stability).

SP-22 Planner

Target tokens/segment

Slider guides segment size (soft). Planner balances timecodes + semantics around this target.

Max segments

Upper bound (UI shows 1–6). Quick mode uses ≤3 to speed runs.

Manual hard splits

Comma-separated HH:MM:SS—planner must cut at these marks.

Replan

Re-runs SP-22 with your latest settings (non-destructive).

Badges

⚠︎ warns if a segment estimates > 2800 tokens (consider extra split or higher target).

Per-Segment

⛏️ Copy SM-22 Prompt

Mines up to 22 high-value quotes (verbatim ≤280 chars) with novelty tags + scores. Paste into your model, run, then collect outputs.

📋 Copy Segment Raw

Copies just the segment text if you want to handcraft or debug.

⛏️ Copy All SM-22 Prompts

Copies one concatenated prompt pack for all segments (fast parallel mining).

SCM-11 Miner

🎼 INIT (S1)

Builds the initial Cognitive Map (Beliefs, Values, Tradeoffs, Assumptions, Vision, Narratives, Collaboration) + Quote Spotlight + QuickJump + NeuroClarity cards. Anti-flattening “UNION-MODE”.

🎼 LAYER (S2+)

Adds only new updates + fresh spotlights. Strict UPDATE discipline preserves signal.

🎼 Aggregate

Compiles all SCM-11 segment outputs into a single worldview + final quote playlist.

🧩 JSONL Compiler

Turns SCM-11 text into GOLD_DATASET_JSONL (nuggets + map_items) + MANIFEST_JSON (no ranks; human-first fields).

Compiler

🔗 Copy Compiler Prompt

Paste all SM-22 JSONL outputs → returns GOLD_DATASET_JSONL, MANIFEST_JSON, and OMITTED_JSONL (deduped & normalized).

🧼 Preflight Repair

Fixes messy JSON/markdown into clean JSONL (drop only irreparable lines).

🧪 Post-Compile Audit

Checks id/timecode format, quote length, novelty set, basic caps/floors; returns a JSON violation report.

Transformers

Universal

Exec Summary, References, Applications, Development (API/RAG). Paste GOLD_DATASET_JSONL where indicated.

Synsian Set

QuickJump (10-min), Primer (20-min + blog), Decision Brief, Sales Playbook, UX Insight, PR Quote Bank, Analyst Brief, RAG Pack.

Copy Buttons

Each card copies a ready prompt. Outputs always cite [nugget_id @ tc] with verbatim quotes.

Copy & Toast

Any “Copy …” shows a green “Copied ✓” toast. If “Open target after copy” is checked, your chosen provider opens in a new tab so you can paste immediately.

Timecodes

Detection

Lines like [HH:MM:SS] anchor boundaries and inform Quote timestamps.

Strong gaps

Planner prefers splits where gaps ≥ ~20s or at explicit timecode starts.

Manual cuts

Exact, hard split at your entered timestamps.

Tokens

Estimator

Hybrid (chars+words) for robust sizing. Target is soft; 2800 is the hard guardrail per segment.

Cap of 6

Intentional simplicity: keeps copy-paste manageable and stable across providers.

Quick mode

Use ≤3 segments for speed; good enough for most drafts.

Workflow

1) Plan

Paste transcript → tweak Target/Max/Manual splits → SP-22 segments appear.

2) Mine

For each segment: run SM-22 (or SCM-11 INIT/LAYER). Save outputs.

3) Compile once

Use the appropriate Compiler (SM-22 or SCM-11 JSONL) to produce GOLD_DATASET_JSONL (+ MANIFEST, + OMITTED).

4) Repair/Audit

Run Preflight Repair (if needed) and Post-Compile Audit for confidence.

5) Transform

Feed GOLD into Universal/Synsian Transformers (QuickJump, NeuroClarity, Briefs, RAG).

Fixes

Oversized segment

Lower target tokens or add Manual split at nearest timecode; replan.

Messy miner output

Run Preflight Repair → clean JSONL → retry compile.

Duplicate quotes

Compiler de-dupes; see OMITTED_JSONL for reasons and kept ids.

Thin segments

Planner auto-merges tiny tails; otherwise raise target slightly and replan.