Three new cards, two new operators. Shopify Engineering published two posts on September 10 that together make the clearest public case yet for how coding agents change engineering platform decisions. Chris Zukowski followed up on September 8 with Steam data that contradicts a widespread assumption about horror game launch timing.
Coding agents changed the platform calculus for mobile
Shopify spent six years on React Native. The reasoning was straightforward: write code once instead of twice. On September 10, the engineering team published two posts explaining why they reversed that decision.
Mustafa Ali's Coding agents erased the cost advantage of cross-platform mobile frameworks names the mechanism directly. The economic case for cross-platform frameworks rests on a single assumption: duplication is expensive. Coding agents break that assumption. Agents handle implementation, translation, testing, and review between platforms. What previously required two full-stack developer workflows now requires one developer directing agents across two worktrees. The per-platform cost premium disappears.
"LLMs changed one of the core assumptions behind our 2020 decision, so we reevaluated our mobile stack from first principles."
Once the cost premium is gone, native wins on every other dimension. Platform-specific capabilities, no abstraction layer to fight, and the freedom to ship the performance characteristics the platform enables.
Jason Kim, Quique Fagoaga, and Max Da Silva's One engineer with coding agents proved a native mobile migration feasible in one week is the companion case study. One engineer used coding agents for one week to port the Shop app's existing React Native screens to native SwiftUI, demonstrating a full feature-for-feature migration was achievable. A team of six then completed the full migration in twelve weeks. The results: iOS startup time fell 23%, Android startup time fell 50%, crashes dropped tenfold, Android app size shrank by 109 MB, and build times fell 75%.
"Agents were particularly effective when they had an existing implementation to work from. They ported defined features, scaffolded screens, wired up data, implemented animations, and refined layouts based on visual feedback."
The workflow that made this work: run multiple agent sessions across separate git worktrees in parallel rather than one engineer, one change, one run. The supervision model replaces the sequential iteration model.
- Mustafa Ali, Coding agents erased the cost advantage of cross-platform mobile frameworks. Coding agents erased the cost premium of maintaining two native codebases versus one shared cross-platform codebase, reversing the economic case that drove Shopify to React Native in 2020.
- Jason Kim, One engineer with coding agents proved a native mobile migration feasible in one week. One engineer using coding agents proved a React Native to native migration feasible in one week; a six-engineer team then completed it in twelve weeks with 23-50% startup time improvements.
Season does not sell games, algorithm does
Chris Zukowski published data on September 8 that challenges a cultural assumption most game developers hold: horror games belong in October.
October is the worst month for serious horror game revenue, not the best shows the empirical result across 7,019 horror games on Steam. October had the highest release volume. For developers earning over $100K, October ranked as the third-worst month for median revenue. June outperformed it. So did several other months with fewer horror releases.
The mechanism is about how game consumption differs from film consumption. A horror film is a two-hour decision made on a whim. A horror game is a multi-week commitment that players schedule year-round. The October spike in horror game activity comes from Steam's algorithm promoting games already in the library or on wishlists during its Halloween sale. Launching into October means competing on a crowded release slate for the same algorithm surface area, not tapping into a seasonal buyer surge.
Only two of the nine most successful horror titles launched in October. Phasmophobia launched in September. Buckshot Roulette in April. The Forest in May.
- Chris Zukowski, October is the worst month for serious horror game revenue, not the best. October is the third-worst month for serious horror game revenue despite having the most releases; game purchase patterns are not seasonal the way film audiences are, and the Halloween Steam sale benefits existing wishlists more than new launches.