Cutting MCP Tools Didn't Make Anything Lighter — 2,377 Bytes of Definitions vs 110,298 Bytes of Response
I wrote a minimal MCP server for a wallpaper catalog and measured the byte cost of tool definitions against the byte cost of responses. Here is which side actually matters, and the real reason to merge tools.
Exactly 20 Characters, Still Rejected: Putting Character Counting Behind One Boundary
Emoji and combining marks make your client and your server disagree about length. I measured four counting methods on Node v22 and found 75.6% of naive truncations split a grapheme and 36.1% produced replacement characters over UTF-8. Here is the shared-module design that fixed it.
Every Experiment Kept Landing on the Same Devices: Hash Choice and Salt Position, Measured Across a Million IDs
On-device experiment assignment looked fine in isolation, then collapsed the moment two experiments ran side by side. Here is what one million IDs revealed about four hash functions, why the culprit was the concatenation order rather than hash quality, and the implementation I settled on, with the measured numbers.
Collapsing Duplicate Requests Into One: A Reference-Counted Single-Flight Layer
When several components fire the same API call at launch, you get a burst of identical requests. Here is a single-flight layer that shares one in-flight promise instead: how to build the key that decides the folding, the trap of handing out a failed promise forever, the trap of one caller's abort cancelling everyone, and the test that keeps it all from regressing, with the real network numbers alongside.
Resolving App Config in Three Layers: Merging Defaults, User, and Remote With Bounded Overrides
A single type-safe layer that merges compiled defaults, user preferences, and remote config. So a broken remote value never takes your app down, each key gets its own override strength, plus schema validation and range clamping, built from a real production incident.
Regenerable Zones in Rork Max Code: Keeping the Freedom to Rebuild
Generated code carries an invisible asset: the option to throw it away and rebuild it. Every hand edit quietly expires that option. Here is how I track it with a ledger and CI checks across six live apps.
Designing Seams That Survive AI Regeneration in Rork
Every follow-up prompt to Rork can quietly wipe out logic you wrote by hand. Protecting it with prompts is a patch, not a fix. Here is how to separate generated code from code you own, and draw a boundary that regeneration cannot reach, with working Zustand and service-layer examples.
Undo That Jumps to the Wrong Place — Designing Edit History You Can Trust
Snapshotting the whole state on every edit crushes memory after a few dozen steps. Here's an undo/redo built on command history plus snapshots, with coalescing, a depth cap, and cross-session persistence, in working TypeScript.
The Device Clock Can Be Moved — Protecting Daily Features and Trial Logic from Time Tampering
Advancing the device clock by one day was enough to claim tomorrow's daily wallpaper today. Starting from that log entry, this article lays out a three-layer time model — wall clock, monotonic clock, server time — and shows how to build daily-reward and trial logic that survives clock rollbacks.
Holding Layer Boundaries in a Rork-Generated Expo App with ESLint and dependency-cruiser
Long-lived Rork-generated Expo apps quietly accumulate screens that import the API client directly. Here is how I froze 214 existing violations as a baseline, eliminated 17 circular dependencies, and made CI reject anything new for 38 extra seconds.
When the Device Runs Out of Space, What Should Your App Protect?
Design Expo/React Native apps that assume writes can fail. Free-space budgets, LRU reclaim, data-protection tiers, and telemetry that surfaces silent failures — drawn from running six wallpaper apps.
Rork Max (Swift) or the Standard Version (React Native): How to Decide as a Solo Developer
Stuck between Rork Max's native Swift and the standard React Native version? Here is a practical decision framework built from a solo developer's perspective, weighing cost, feature boundaries, and how easy each path is to migrate later.