Sonnet 5.5, GPT-6.1 Sol, or Opus 5.5? I Decide by the Kind of Fix Before I Pick the Model. A Week of Logging Where My Rork Credits Went
After Rork's model menu settled into Sonnet 5.5, GPT-6.1 Sol, and Opus 5.5, I sorted my fixes into three kinds and assigned a model and effort level to each. Here is how I read the changelog, the tiny ledger that tracks where each credit goes, and the rule I drew from a week of entries.
Your Expo Router App Flashes the Sign-In Screen on Every Launch. Rebuild the Guard With a Third State: Not Known Yet
Signed in, yet the sign-in screen flashes at launch; signed out, yet the screen never changes. How I rebuilt Expo Router Protected routes around three Supabase auth states, including the one the docs never name, and where the _layout.tsx files have to live.
Keep Fixing in the Same Thread, or Restore and Rephrase? What I Learned Fixing One Rork Bug Both Ways
When a Rork fix request fails for the third time, do you keep going or use Restore and ask differently? I fixed the same bug both ways and compared round trips, leftover code, and where I now draw the line.
Building the Same Gallery Grid with Opus 5.5 and GPT-6 Sol in Rork — Choosing a Model by Its Effort Steps
Rork's model menu now lists Opus 5.5 and GPT-6 Sol. I built the same three-column gallery screen with both, compared their effort settings, and kept the code I'd actually ship. Here is what differed and how I chose.
A build that stops with exit code 0 — the flag that silenced the log, and the check that let an empty file through
Two reports landed in the same week: eas build --local stopping with exit code 0 but produced no further output, and a zero-byte stub that never got embedded. Both came down to settings that reduce output. Here is how to audit your own scripts.
Two logins to check first when Expo Go stops opening your preview
When the QR code stops loading your Rork preview on a real device, check the logins before the router. Here is which path needs a sign-in on both ends, and which paths are untouched.
Reading only what Rork changed: the diff routine I run after every re-export
How to read a fresh Rork export as a diff against the last one: clear tracked files before you copy, strip the noise before you count, and stop once before the merge commit lands.
The three lines I check when prebuild stops on the standard SDK 57 Swift AppDelegate error
Turning on ios.enableSceneSupport for iOS 27 can stop prebuild cold. The cause was not my SDK version but the shape of AppDelegate.swift. Here are the three lines to check, and why copying the SDK 58 migration steps by hand will break your build.
The recommended replacement was already gone — auditing all 74 rows of Gemini's deprecation table
Cross-checking Gemini's deprecation table turned up five rows where the recommended replacement had itself already shut down, including gemini-2.5-flash-image, which goes away on October 2. Here is how to verify a migration target, and how to catch the 404 before your users do.
The Token I Deleted Came Back After a Restart — Detecting Failed SecureStore Deletes
On Android, getItemAsync can report null right after sign-out while the value is still sitting on disk. Here is how I rewrote a sign-out path that was verifying deletion by reading instead of by the delete result.
Two Balances, Not One: Reading Rork's Build Credits and Cloud Credits Apart
Rork keeps two separate balances. Here is how to tell which one a charge came from, why one prompt can cost more than one credit, why a Cloud balance that never moves may mean your AI is not running at all, and what changes once real users start spending it for you.
Where the AI decision lives once Rork splits your app into two native codebases
New Rork apps are Swift on iPhone and Kotlin on Android — two separate codebases. Here is how I decide which AI-related judgments stay in each client and which move to the shared backend, plus the minimal contract both clients read the same way.