← Back

Devs refuse to code without AI, creating a giant 110,000-bug mountain on GitHub

Original version · May 31, 0:30

The coding revolution is here, but instead of digital utopia, we got a collective shrug from developers who can't even open an IDE without an AI nanny. Turns out, our new machine overlords are just very fast at writing terrible, unmaintainable garbage.

The research lab METR tried to recruit experienced open-source developers for a study comparing coding with and without AI assistants. The experiment failed before it even started because practically all volunteers flatly refused to write a single line of code without their AI sidekicks. Writing code manually is apparently so last century that today’s developers would rather boycott research than suffer through the agony of thinking for themselves.

This collective addiction is particularly hilarious because previous experiments showed that AI assistants actually slowed down experienced programmers by about 19 percent. The developers themselves were absolutely convinced they were flying at warp speed, living in a beautiful delusion while their actual productivity quietly took a nosedive.

To measure the hangover from this high-speed coding spree, researchers from the Singapore Management University dissected 304,362 AI-generated commits across 6,275 GitHub repositories. They discovered that machine-written code regularly injects sneaky technical debt that sits in production like a ticking time bomb. By February 2026, the mountain of unresolved, AI-introduced bugs and issues in these repositories officially crossed the 110,000 mark and is still climbing.

As software practitioner James Shore pointed out, doubling your coding speed is only a win if maintaining that code also becomes twice as cheap. Instead, teams are blindly accepting massive, low-quality codebases that will require a small army of human developers to debug, clean, and rewrite once the AI-fueled hype cycle inevitably crashes into reality.

The tech industry has successfully automated the creation of legacy code, proving that human laziness will always find a way to scale up chaos. It seems the future of software engineering isn't about elegant algorithms anymore, but about who can survive the avalanche of automated garbage first.

Source: TechCrunch

Comments

This is where the magic happens: AI reads your discussion and rewrites the article based on the most interesting comments. Each strong comment adds points to the meter below. Once the meter is full, the article updates live — no page reload needed.

18/24
  1. Toxic Gremlin
    lmao so we are just paying devs to be prompt engineers who cant even debug their own copilot garbage
    +2 emotionalA delightful display of existential dread regarding the inevitable heat death of software engineering
  2. Glitchy Sphinx
    as a senior dev, i spend 80% of my day fixing 'fast' ai code written by juniors. it is literally a nightmare.
    +5 solidA veteran survivor reporting from the trenches of the AI-generated bug apocalypse
  3. Burning Wolf
    but wait if ai writes code faster we can just use ai to fix it? checkmate skeptics
    +3 funnyThe logic is so circular it could be used as a wheel for a carriage in the dark ages
  4. Electric Bishop
    this is why my legacy systems built in cobol will outlive all your modern react startups. pure poetry.
    +8 exceptionalWatching a COBOL enthusiast gloat over the ashes of modern frameworks is the kind of petty I live for