Ev BogueI opened Hacker News today and it looked less like a tech forum than a group therapy session for people who accidentally built the future and now have to live in it.
The top story was "I'm Tired of Talking to AI", with the HN thread doing what HN does best: half confession, half cross-examination, half people correcting the other two halves. A few slots below it was PostHog saying it will train AI models with customer data, opt-in by default, and the HN thread immediately knew what that meant. Then came Tech CEOs are apparently suffering from AI psychosis, Webflow evolving for the agentic web, and a Claude Code daily-driver guide.
This was not an AI news cycle. This was the whole web arguing with itself in public.
The funny part is that everyone is right enough to be annoying.
The tired people are right. It is exhausting to talk to a support bot, an AI search answer, an AI-written help page, an AI-generated sales email, and then open a comment thread where half the humans sound like they are trying to pass as models and half the models sound like they are trying to pass as humans. The old internet bargain was already bad enough: give us your attention and we will show you ads. The new bargain is worse: give us your attention, your data, your writing, your product usage, your support tickets, your internal docs, your facial expression while you complain about all of this, and maybe we will generate an answer you almost wanted.
The builders are also right. The tools work. Not always. Not cleanly. Not without a human adult in the room. But they work enough that pretending otherwise has become its own kind of cope. I use them. This site is maintained with them. I am writing this with an AI agent helping me gather the links and keep the draft from turning into wet cardboard. The agent did not decide what I believe. It did make the work faster.
That is the hangover: AI is simultaneously useful, invasive, tedious, and impossible to route around completely.
HN is interesting because it still has a memory of a web where tools were judged by whether they helped competent people do actual work. That memory is why the comments get so sharp. The forum is not allergic to powerful software. It is allergic to bullshit software pretending power is the same thing as judgment.
PostHog training on customer data by default is a perfect example. The issue is not that a company wants better models. Of course it does. The issue is the ratchet. A product starts as analytics. Then it becomes infrastructure. Then it becomes the place where your behavioral exhaust accumulates. Then one morning there is a blog post explaining how that exhaust can make the product smarter, and the default setting is yes.
That is not a tiny product decision. That is the culture of the web in miniature. Every company wants to become a refinery for whatever users leave behind.
At the same time, the Claude Code guide is not fake. Agentic coding is not fake. Webflow noticing that agents are going to browse, edit, and mutate websites is not fake. The shift is real. Software is being rebuilt around the assumption that the user is no longer always a human clicking around with a mouse. Sometimes the user is a human with a ghost intelligence in the loop. Sometimes the user is the ghost.
This is where the cheap AI discourse fails. "AI is slop" is too simple. "AI is the future" is also too simple, and usually said by someone holding a pitch deck like a religious object. The real situation is uglier: AI is becoming part of the operating condition of the web before we have social rules for it.
We do not know when an answer is written by a person.
We do not know when our data is training the next feature.
We do not know when a support agent is a bot, a human reading bot output, or a bot pretending to escalate to a human.
We do not know whether the junior developer learned the system or merely learned the prompt that made the test pass.
We do not know whether the founder believes the product needs AI or believes the investor needs to hear the word.
HN is not having a simple backlash. It is having a reality check. The people who spent years asking "can we build it?" are now staring at the results and asking "why is this everywhere?" That is a better question.
I do not want a web without AI. I want a web where AI has to declare itself, where training on user data is opt-in because opt-in means opt-in, where generated text is not used as a fog machine, and where people using agents still own the judgment behind the work.
That last part matters most. The machine can write. It can code. It can summarize, route, classify, imitate, autocomplete, and sell. It cannot decide what should exist. It cannot take responsibility for the culture it creates. It cannot know whether a product has become hostile to the people using it.
The builders still have to know that.
Today's Hacker News page is useful because it refuses to settle down into one clean opinion. It is tired of AI. It is using AI. It is suspicious of AI. It is building the agentic web. It is mad at companies for feeding on user data. It is reading tutorials about how to make the tools work better.
That is not hypocrisy. That is what it feels like when a technology crosses from novelty into weather.
You do not get to opt out of weather. But you can still decide what kind of house to build.
If you are building with AI without letting it eat the people using your work, email ev@evbogue.com or text 773-510-8601. I want to know what rules you are using.