What reversing, modernising old games tells us about the economic impact of AI

A recent reverse‑engineering project rebuilt War of the Lance (1989) for modern browsers by using Qwen-3.8-Flash-Next to extract rendering and rules and port them into a Three.js renderer, then iteratively modernised the experience toward a Total War/Heroes style. The workflow combined code synthesis, headless Blender automation generated by models (Fable 5.1, later GPT6 Astra) to produce web-ready GLBs, and an automated harness that enforces linting, testing and autonomous playtesting. Concrete metrics: a faithful port required roughly a week with a single local lane and under four hours of focused human attention; cloud-parallel lanes would cut that to days, recent frontier models to hours, and projected future models to minutes. Small asset tweaks like seasons/day-night cycles took minutes and cents to implement, and an eight‑hour Fable run produced an ~85% working DragonStrike port.

The broader argument is that transformer models are commoditising deep, multidisciplinary knowledge work central to game reverse‑engineering and creative production, turning skilled engineers into “bot herders” who provide intent and curation rather than hands‑on execution. This shifts economic leverage: models scale knowledge application cheaply, eroding the scarcity that underpins wage premia and raising IP and regulatory challenges as enforcement lags. While new roles may emerge, reliability, copyright and governance bottlenecks are unlikely to prevent long‑term disruption; centralised model providers with economies of scale will outcompete many traditional knowledge workers unless policy or market structures change.