AI Is Causing Signal Collapse Everywhere All at Once
The author argues that AI makes cheap to fake the signals we use to judge quality, such as GitHub stars, polished writing, and homework passing autograders, so good work stops getting rewarded and a market for lemons forms. He uses signaling theory to explain why this change is so fast and disorienting compared with earlier, slower erosion.
- Shortcuts like stars, clean READMEs, and passing tests worked because faking them used to cost nearly as much as doing the real work; AI removes that cost gap.
- Buying GitHub stars already cost about $0.07–0.20 each, and AI now makes fake accounts with realistic histories easy to produce, hiding the fakery better.
- When signals collapse, evaluation gets more expensive, and buyers, employers, and reviewers stop paying a premium for good work, so poor work dominates, as in Akerlof's 1970 used-car lemons paper.
- Past signal erosion (biological mimicry, CAPTCHAs, pass/fail grading) was usually slow, but AI is breaking many signals across open source, education, and research at once.
Signal collapse is the author's name for what happens when AI makes the cheap cues we use to judge quality easy to fake all at once. Humans have always relied on shortcuts: a package with many GitHub stars, a student whose code passes an autograder, a paper with a tidy related work section. These work because faking them used to cost nearly as much as doing the real work. The author draws on signaling theory and the 1970 "market for lemons" paper by Akerlof, which describes how buyers who cannot tell good goods from bad will stop paying for quality, and good goods then disappear. The key difference he stresses is speed. Grade inflation or spoofed caller IDs eroded signals over decades. AI is collapsing several at once, across open source, education, and peer review, because they share a common root: writing, coding, and interacting are exactly what generative models now produce cheaply.
The examples are concrete. Stars on GitHub already could be bought at roughly $0.07 to $0.20 each, but AI now helps build fake accounts with plausible histories. A polished README, auto-generated issue replies, and a clean pull request with passing tests can all be produced without any real understanding of the project. In education, an essay or assignment can be produced with no learning involved, and in peer review a convincing related work section can be generated without the author having read a single cited paper. The author is careful to say that the old shortcuts were never fully reliable. Polished writing favors native speakers, and a clear README may reflect free time more than code quality. Still, the shortcuts were what let these systems operate at scale, because full evaluation of everything is impossible.
The practical consequence is the lemons problem. If you cannot cheaply tell good work from bad, good work stops getting rewarded, and poor work comes to dominate. Reviewers who suspect a paper took minutes to generate will not spend hours giving feedback, and employers have less reason to pay for graduates whose credentials no longer say much. The author argues we will need new signals or accept higher evaluation costs, but only where the benefit of a better decision outweighs the expense of checking. He is clearer about the diagnosis than the cure, pointing toward signaling-theory vocabulary as a starting point for discussing fixes rather than offering a finished answer. For anyone who maintains open source, grades students, or reviews papers, the piece is a useful framing of why the usual gut checks have started to feel unreliable.