AI coding's real tragedy: software dies, but who gets to create?
A translated commentary (by Ian Bogost, originally in The Atlantic, republished by Huxiu) uses the claim that Adobe software was reverse-compiled by AI and released as open source to frame Silicon Valley's argument over whether software is now dead or worthless, with indie developer levelsio and others declaring the moat gone. Drawing on a year of building small apps like Beanslaw and Taptimer with AI, Bogost argues that AI coding does restore ordinary people's ability to write software for themselves, but that the most rewarding work comes from a back-and-forth partnership — you supply intent, the model implements, you judge and revise — rather than one-prompt delivery. He notes that agentic tools such as Claude Code and OpenAI Codex push toward full automation, while Apple's $99-a-year developer program and platform rules over things like haptics and distribution risk handing the creative process back to a few large companies; the Adobe episode is cited only as background to the debate, without verifiable detail.
It reminds practitioners that AI coding's value depends not just on generation speed but on who controls platforms and the creative process.