You type your deepest academic anxieties into an AI planner. Where does that go?
Most apps: Straight to their servers. Analyzed. Stored. Possibly sold. Used to train models. Subpoenable.
Locked In: Saved in your browser. Syncs only if you explicitly sign in.
What "local-first" actually means
Default: saved data stays local
- Open app → loads from cache → works offline
- Type chaos → AI processes via API (only prompt + context) → plan returns
- Plan saves to
localStorage(lockedin-v3key, migrated fromlockedin-v2) - Mood check-ins, timer sessions, streaks — all local
- Your saved data stays on your device unless you sign in
Optional: account sync (you control)
- Sign in with Google or email → then Supabase sync activates
- Data travels over HTTPS to your own account
- Row Level Security on every table —
auth.uid()only,anongrants revoked - You can sign out → local data remains, cloud data untouched
- Clear all → wipes both local + cloud
What the AI sees (minimal context)
Each request sends only:
- Current message + last 10 messages
- Mode (Study/Chill/Crunch/Roast)
- Effort level (Low/Mid/Max)
- Aggregated stats — streak, completion rate, avg mood, preferred modes
- Your last 5 plan outcomes and active goal titles (so plans adapt and aim)
- Never: your name or email, the raw mood timeline, calendar events
Why this architecture is rare (and hard)
| Challenge | Most Apps | Locked In |
|---|---|---|
| Offline support | ❌ Requires connection | ✅ Full offline after first load |
| Conflict resolution | ❌ Server wins | ✅ Local wins, cloud merges |
| Encryption | ❌ TLS only (server decrypts) | ✅ TLS in transit; most data never travels at all |
| User control | ❌ All-or-nothing | ✅ Guest → signed in → guest seamlessly |
| Data portability | ❌ Export requests | ✅ Share card, ICS, copy JSON |
Real privacy threats this prevents
| Threat | Cloud-first app | Locked In |
|---|---|---|
| Data breach exposes your anxiety, grades, habits | ✅ Vulnerable | ❌ Not there unless you opt into sync |
| Company sells analytics to third parties | ✅ Common | ❌ No data to sell |
| AI training on your personal content | ✅ Often in TOS | ❌ Only aggregated stats |
| Government subpoena for user data | ✅ Must comply | ❌ Only ever your opted-in account rows |
| Employee access to your journals | ✅ Possible | ❌ Requires your signed-in account (RLS) |
| Company pivots/shuts down → data lost | ✅ Risk | ❌ Local copy always yours |
The "but convenience" argument
"But sync across devices is convenient!"
Locked In gives you both:
- Guest mode — everything works locally, 5 free AI messages to start, zero friction
- Sign in — sync to your own account, same features, you decide when
- Sign out → back to guest, local data intact
No "create account to continue" walls. No "premium for sync." You own the data flow.
Technical details (for the skeptical)
Sync: HTTPS (TLS) in transit; rows stored in your own Supabase account behind Row Level Security
auth.uid()::text = user_idon all 4 policies (SELECT/INSERT/UPDATE/DELETE)anonrole: ALL GRANTS REVOKED — zero anonymous access- PKCE auth flow (no implicit flow token leakage)
Storage key: lockedin-v3 in localStorage (existing lockedin-v2 data is migrated once, old key kept as backup)
- Migrates legacy
sb-*-auth-tokensessions automatically consumeSignedOutFlag()prevents immediate re-redirect after sign-out
Try it — verify yourself
- Open lockedn.in in incognito
- "Start free" → "Continue as guest"
- Type something personal
- Get plan → complete it → check streak
- Open DevTools → Application → LocalStorage →
lockedin-v3 - See your data. See NO network requests to Supabase.
- Close tab → reopen → data persists
That's local-first. Your chaos, your device, your choice.
Locked In: Free, local-first, private-by-default AI productivity. Try it here — no account required. Who builds it and why.