Sniffs the cost of outrage-bait & engagement traps.
Mute weaponized UX and algorithmic pokes.
Verify “offline” claims with a 3–10 min test and receipts.
Separate brand lore from facts you can cite.
Time-stamp a claim so it can be wrong later (on purpose).
Link evidence across posts without losing context.
Quick “does this touch the real world?” screen.
Keeps projects from bloating into feature soup.
Turn vibes into policies with failure conditions and dates.
Strip weasel words; force responsibility → harm → remedy.
Scores repetition, buzzwords, cliché foam, and rhythm.
Catches skeleton screens & placebo spinners.
Find claims that need sources; export a receipt demand list.
Boils release notes into what changed and why it matters.
Flags template-y AI stock vibes: symmetry, noise, sameness.
What lives here
Real problems, fixed in real time: what this fixes → exact steps → how to verify → how to undo.
Vendor-agnostic whenever possible. Proof over faith. Lowest permission wins.
You don’t need a “definitive guide.” You need the switch to flip, the leak to stop, the nonsense to shut up.
Switchcraft is a shelf of short, testable playbooks—privacy toggles, billing brakes, local models, sanity checks. If it takes longer to read than to do, it doesn’t ship.
Got a mess we should fix? Send steps + a screenshot via Contact (subject: Switchcraft). If we can reproduce it, we’ll publish—credit if you want it, quiet if you don’t.