Bun's million-line AI rewrite is the strongest evidence yet for 'the end of programming'
InfluxData cofounder Paul Dix picked the wrong week to be ignored. His essay "The End of Programming" argues that hand-written, human-reviewed code is headed for extinction — and his central exhibit is Bun 1.4, a release whose own engineering blog confirms almost every number he leans on.
InfluxData cofounder Paul Dix says the act of writing code by hand and having other humans review it is headed for extinction — and he rests the argument on how Bun 1.4 got built. Writing on his personal blog, Dix points to the JavaScript runtime's rewrite from Zig into Rust: roughly a million new lines of Rust produced by parallel AI agents working for 11 days under one developer, Jarred Sumner. Bun's own writeup puts it at over 6,500 commits driven through about 50 continuously running agent workflows in Claude Code, using a pre-release version of Anthropic's Fable 5 model — compute that Dix estimates would run about $165,000 at API prices. The result now ships to millions of developer machines.
The part worth dwelling on is not the speed. It's what Sumner did instead of reading the diff. Nobody can review a pull request that adds a million lines, so Bun built its confidence differently: a language-independent test suite with around a million assertions, adversarial code review performed by other AI models whose only job was to find how the change is wrong, and — most telling — a rule of fixing the process that generates the code rather than hand-fixing individual bugs. That is a different software discipline, not just a faster one. Verification replaces inspection; you trust the loop because you've tested the loop.
Dix pairs this with his own experiments as InfluxDB's creator: he gave Fable an architecture for an Apache Iceberg integration and came back 14 hours later to a working implementation, then let an edge-replication feature run 28 hours to a mostly functioning build that is still operating months later. His forecast follows directly: OpenAI's next model lands within weeks, frontier-level intelligence gets cheap enough for constant agent armies within a year or two, and we end up with more production software written by machines than by humans. There will also be, in his words, a mountain of slop — and GitHub's outage post showing an exponential climb in pushed code since last year suggests the deluge has already started.
Our take: the essay oversells the timeline but undersells the mechanism. The org-chart reality is that most companies still require humans to own every line, so traditional programming will persist for years in exactly the places that matter most. But Bun 1.4 quietly settles the deeper question — machine-authored code has already shipped at infrastructure scale, held together not by human eyes but by verification systems no human fully reads either.
Separately, Bill Gates told the New York Times that tech executives privately admit they are "very worried" about AI-driven disruption while publicly downplaying it — because candor would threaten fundraising and planned IPOs. The interview lands alongside his GatesNotes essay warning that the AI era will be among "the most turbulent times in human history" and that "we are not preparing adequately." We covered the essay itself this morning — Bill Gates says AI has crossed its danger thresholds — now he wants a plan — but the new claim is sharper: it accuses an entire industry of coordinated risk-washing in service of the capital markets. Gates proposes reserving certain jobs for humans and wants to discuss AI governance frameworks with China's leadership. If even the optimists' optimist thinks founders are managing their IPO narrative instead of the risk, expect that quote to surface every time a lab downplays a safety report before a funding round.
What to watch: whether Astra, expected next month, makes agent-scale coding cheap enough that Dix's "year or two" compresses into quarters.
When your own job is the thing being declared extinct — do you believe him? Tell us in the comments.
