19 stories
404 Media says it is inundated with AI-written emails from people with ordinary tech problems. After it began asking senders how they found it, every one it asked said a chatbot recommended contacting it.
ChatGPT, asked how to recover a banned Instagram account, concluded that emailing Meta was probably hopeless.
The top HN threads: facts from the discussions, both sides, insider voices. Email, Telegram, or RSS.
Metehan Yesilyurt of Peec AI built about 50,000 pages of invented stats plus 15,000 listicles with GPT-4.1. Google demoted the site after two updates, but ChatGPT kept crawling it, logging over 600,000 bot requests in 30 days.
The fake figures reached an academic paper and investment research from Qatar's largest financial group.
The preview supports Ubuntu 24.04 and 26.04 LTS, Debian 13, Fedora 43 and 44, on x64 and ARM64 through deb and rpm packages.
Early users immediately tested the boundary: Wayland input methods, downloads from another OS, and whether desktop projects can see existing CLI work.
OpenAI announced ChatGPT updates to GPT-5.6 Sol and broader free-user access to GPT-5.6 Luna.
HN debated whether widening access reflects competition or a more useful product principle: ordinary users should not need to choose a reasoning setting before asking a question.
Terence Tao published a raw ChatGPT transcript exploring a counterexample to the century-old Jacobian Conjecture — not a summary of the math, but the actual back-and-forth, phrased like an expert cutting straight to the point.
One line stood out to readers as unmistakably his register: "I've activated Pro. Can you continue to look for a potential geometric explanation of the X_3 ~ A3 miracle that avoids coordinates or other unmotivated constructions?"
Someone tried the trick everyone eventually tries: ask ChatGPT to guess who it's talking to from the same transcript. Unprompted, it guesses "math researcher or paper author."
The page that just went live promises ads "clearly labeled" and "separate from answers," rolling out first to only seven countries, about 8% of the world's population, and skipping every paying subscriber.
The asterisk sits in OpenAI's own targeting rules: sensitive chats stay ad-free, but the memory they leave behind doesn't. It can still steer which ad shows up in an unrelated chat the next morning.
Her blog's URLs are date-stamped, and the phantom hits match that pattern too closely for coincidence: a model invents a post, hands a reader the link, and the link goes nowhere. She frames it as Michael Crichton's "Gell-Mann Amnesia effect": catching a source's confident errors on a topic you know, then trusting it on everything else.
A well-known HN security engineer dismissed her argument as relitigating a two-year-old debate. A second blogger said the same pattern hits her own site, her name attached to things she never wrote.
The detector flagged a 2015 conference paper at 74% machine-written and a 2012 PhD dissertation at 40%, both submitted years before large language models existed, according to a commenter who ran his own decade-old work through the tool.
The tool's creator conceded in the thread that he has no clean way to build a control group of human writing from after 2023, since every available shortcut relies on the very detectors he's trying to outperform.
Hotz's essay dismisses the idea that a smart-enough AI beats physical reality, then mocks a rival AI-safety plan for illustrating its case with a render of an ocean datacenter. His proof today's AI is broken: a screenshot where ChatGPT tells him to call 911 instead of helping him hide a body.
HN's read split: replies mocked the murder test as a bad look even from people who share his conclusion, while others ran his line about tokens never taking over the world into a counterexample, dictators who conquered nations with nothing but words.
Grauer's essay names the moment exactly: a mentor she'd built up over years waved her toward a chatbot instead of sharing the judgment he'd use himself. She says she'd already asked twice.
The 105-comment thread split between engineers who called the redirect disrespectful and others who defended it as an earned shortcut, until one open-source maintainer admitted Claude would likely nail 98% of his own advice anyway.
The satire tracks 2022–2026: each skeptic goalpost — Turing test, Snake game, one-person unicorn — lasts about 18 months before it falls, set against booster claims like Google's 'quarter of new code is AI' and a founder's $1M ARR in 17 days.
The thread's flip is that the page mocking skeptics is itself AI slop that fabricates its own evidence, and that the real grievance was never capability — it's how executives treated workers on the way.
A Brown economics professor who studies game theory gave a take-home, closed-book midterm on the honor system, then a proctored final. The grade distributions diverged so sharply he says at least 50 of 86 students cheated with AI — and his own department shrugged.
The thread was brutal: a game-theory expert who can't foresee that everyone defects built the trap himself. Meanwhile a Dartmouth instructor is rebuilding a CS course as an "adversarial problem" against students optimizing grade-per-effort.
Trakkr ran six models through 61 political questions, twelve times each, web search and reasoning off to mimic the consumer default. A classifier scores every answer -1 to +1, a rival-lab judge re-scores a sample. Gemini lands neutral, ChatGPT at -0.29, Grok at +0.21.
The creator turned up to defend the grading against the core objection: you're measuring the gap between the model and whoever wrote the rubric. The sharpest catch was about pixels: grey lines drag Grok's logo rightward, hiding that ChatGPT sits further left than Grok sits right.
The single em-dash tell is dead, the author argues. What gives AI away is aggregate sameness across many outputs, not any one artifact.
His proof: a collage of roughly 150 AI-generated kids' encyclopedia covers, many of them category bestsellers, recycling the same tropes and surnames.
The thread named the mechanism, then questioned whether the sample proves anything.
A blog post titled 'The AI Hate Progression' traces one person's slide from amused indifference to total rejection, built on a single grievance: 'Consent is just gone. It does not exist when AI enters the room in 90% of cases.' Its ask: start over, with consent baked in.
The thread shared the anger and demolished the prescription. tptacek's read: 'start over from zero' sounds utopian enough to give him permission not to care. Others noted consent died with web indexing in 1999, and that the real target is hypercapitalism wearing an AI mask.
The cynicism was specific: cut prices now, IPO while the hype holds, and the comment everyone nodded at — how long can you keep selling a dollar for 50 cents?
A genuine market split surfaced too: developers prefer Claude for coding, while ChatGPT owns the mainstream. A price war defends the consumer base the rival isn't really attacking.
The post revisits OpenAI's Feb 2019 decision to withhold the 1.5B-parameter model, the nine-month study that found no real misuse, and the eventual full release that November.
The split is sharp: one camp says the social damage from cheap synthetic text is now astronomical, the other says it was a 'we sure hope no one takes us seriously' marketing move.
The synthesis came from a data scientist who built GPT-2 tooling: justified caution in 2019, marketing gimmick when applied to 2026's models.
The argument from the top: if AI were displacing workers, openings would collapse and unemployment would climb. Instead there are more jobs than workers.
The counter from people who hire: strip out healthcare and most sectors are flat or shrinking, junior software roles are near zero, and a chunk of 'openings' are ghost jobs that cost nothing to post.
A line from a Google engineer that stuck: you almost get laughed at asking for headcount now.
43.8% is not a quality score — it's a census. Listicles dominate AI citations for the same reason they dominated Google search: they make up most of the web, and that is what the models trained on.
In the same thread, one reply put a deadline on it: 'six months until the chatbots start penalizing the pattern.' SEO's two-year trap, new platform.
