The anti AI manifesto that barely moved Hacker News

A piece titled "My Anti-AI Manifesto" hit Hacker News on 26 August 2026 and stalled at 2 points and 1 comment. No front page, no four hundred reply thread, no newsletter pickup. The original lives on a personal page hosted on Altervista, the free Italian host many of us associate with the early 2000s web, and its Hacker News submission is still practically empty.

That number is the part we find relevant. Two years ago, a manifesto against generative AI signed by a developer with no prior audience would pull several hundred points and a full afternoon of argument in the comments. Today it competes with dozens of similar pieces published the same week, and with a reader who has already read every argument in it. The genre has not disappeared: it has saturated.

The anti AI manifesto now has a fixed shape

Anyone who has followed the debate recognises the structure without opening the link. There is usually a section on authorship and consent in training data, another on the quality of generated code or text, a third on the energy and water cost of data centres, and a closing passage on the erosion of the craft. There is almost always a personal declaration of refusal: I will not use these tools, I will not publish generated material, I will not take work that requires it.

The shape matters because it explains the metric. Once a format stabilises that much, it stops carrying new information and starts working as a membership signal. The 2026 Hacker News reader already knows what paragraph four will say, so they neither vote nor comment. That does not invalidate the arguments, but it does change where they take effect: they have left the aggregator and moved into contracts, internal company policy and the terms of freelance platforms.

Why this matters to people who do integrate AI

We build Claude integrations, MCP servers and custom agents, so the expected move would be to ignore these texts. We do the opposite, for a practical reason: a manifesto's list of objections is, almost literally, the list of questions that comes up in the first meeting with a hesitant client. Where does the data come from, who answers when the model gets it wrong, what happens to the person who used to do that task, what does inference actually cost once the pilot becomes production.

A team that walks into that meeting without concrete answers loses the project. And concrete answers cannot be improvised: they are architecture decisions taken earlier. Putting in writing which data leaves the client's infrastructure, bounding with explicit permissions which tools an agent may call, logging every call so it can be audited later. None of that is solved with enthusiasm.

What the point count does not tell you

It is worth not reading those 2 points as a verdict on the quality of the text. Hacker News rewards submission timing, headline and the luck of the first vote as much as content. A manifesto posted on a Wednesday morning in August, from a domain with no accumulated reputation, had few options regardless of what it said inside. We have seen considerably worse pieces reach the front page with a better placed headline.

What the number does indicate is the state of the conversation. The argument about whether to use generative models at all has become unproductive in general purpose forums, while the argument about how to use them with guarantees is still alive in more specific places. That is a healthy shift, even if it is uncomfortable for anyone who expected the debate to settle on a single position.

Who benefits from reading it

Product leads about to ship a model based feature who need to anticipate internal pushback before it arrives. Freelancers deciding what to put in their terms about AI use. And anyone writing on the topic who wants to avoid repeating a format nobody reads any more.

Our position is measured and has not changed over the past year: most objections in these manifestos are reasonable, and they are answered with engineering and with contracts, not with proclamations in either direction. That this particular one went unnoticed says more about the exhaustion of the format than about its author.