Everything in one place. This time with real numbers.
Same cockpit. Real data this time. The top 10 questions are mined from 670 actual messages across your Claude Code sessions. The concepts are ranked by how often you bring them up — not by my guess.
One-sentence daily intention · anchor task · 3 tasks · 7-day streak · pitch page for the widget concept.
Top 10 Questions (Real)
Mined from 670 real session messages
What's different from v1
v1 had questions I invented based on general engineering wisdom. v2 has the questions you actually asked, pulled from your Claude Code session JSONLs. Each card shows the 5th-grade version, the verbatim quote, why it matters, and how often it appeared.
01
Where did you save it? Send me the link.
You actually wrote: "where are they really with a real link"
Most-mined ask: link-or-it-didn't-happen. The dashboard memory rule literally requires file:// links on every output.
Mentioned in 65 messages
02
Did we finish ALL of them or just some?
You actually wrote: "Do we have 16 changes for 16 shiner files"
Completeness check. When you run a batch (16 songs, 30 posts, 46 conversations), the next question is always: did we miss any?
Mentioned in 89 messages
03
Show me with a link I can click — every time.
You actually wrote: "when i need to review always link to it"
Stated as a rule, not a question. Already a permanent feedback memory.
Active feedback memory: feedback_end_responses_with_links.md
04
Don't save it in the wrong folder.
You actually wrote: "can you not save this in the radiohead folder it's so confusing because its not radiohead"
Folder-correctness ask. The 'don't conflate unrelated things in one bucket' pattern.
Mentioned in 39 messages
05
Can the same tool work on different stuff?
You actually wrote: "can you do this same with mp3 drum and mp3 bass tracks?"
Generalization ask. If a skill works for one input type, can it work for the cousin? Extension over reinvention.
Mentioned in 18 messages
06
How much did this cost in tokens?
You actually wrote: "what was the usage for this last 2 midi run Shiner-Analysis"
Token-budget question. Already a permanent feedback memory.
Active feedback memory: feedback_token_warning.md
07
What might break if we do this?
You actually wrote: "will midi-analyze skill need fixing"
Prevention question. Before changing X, ask what else touches X.
Mentioned in 5 messages
08
Where's the rules written down?
You actually wrote: "whats the doc that has the methodoology"
Where-is-the-source-of-truth question. The premise behind FILE_TRACKING.md, dashboard_usage.html, every README.
Mentioned in 11 messages
09
Can we even change this thing?
You actually wrote: "Are there any customizations possible to the finder browser and folder icons"
Capability-discovery question. Before designing around a constraint, find out if the constraint is real.
Mentioned in 8 messages
10
What is this file? I don't recognize it.
You actually wrote: "what is this MagRun1-validation.md at file://..."
Identification question. If a stray file shows up, ask before deleting.
Mentioned in 3 messages
Concepts Worth Keeping (Ranked by Real Use)
Ordered by how often you mention them
What's different from v1
v1 had concepts I picked from your memory file in the order I thought they were important. v2 ranks them by how often you actually invoke them across 670 real session messages. The bars are accurate.
01
Rate everything so I know what's good
Evaluate-and-Score
Found in 225 of your 670 session messages (33.6%)
02
Leave a note before you move anything
File Tracking Ledger
Found in 170 of your 670 session messages (25.4%)
03
Save the old one before you fix the new one
Duplicate-Before-Changing
Found in 161 of your 670 session messages (24.0%)
04
Write it, hate it, fix it (automatically)
Self-Critique Loop
Found in 153 of your 670 session messages (22.8%)
05
Use the cheap brain first, expensive only when stuck