Hacker News Daily: Top Tech Debates & Stories You Can't Miss

Large language models are hitting eighty to ninety-five percent accuracy on coding tasks. It's the perfect follow-up to our conversation yesterday on the gap between AI hype and the ground truth, because that number — eighty to ninety-five percent — is either a miracle that will change software forever, or it’s a time bomb. And the entire debate on Hacker News today, May twenty-sixth, two thousand twenty-six, is about which one it is. The conversation is being called “The Eternal Sloptember,” and it’s this idea that we’re about to be flooded with code, and text, and ideas that are almost right. Good enough to work, but not good enough to trust. It’s a crisis of quality.

And that crisis of quality is the thread connecting everything today. Because while developers are arguing about AI-generated code, the state of California just backed down from a law that would have forced open-source projects like Linux to implement age verification. The community backlash was immense. They essentially argued that you can’t treat a global, decentralized open-source project like a single company. The interesting part, though, wasn't the backlash—it was the proposed solution. A purely technical one: a new web header, RTA for "Restricted to Adults." The idea is that browsers would block any site that doesn't have a valid header by default.

It flips the model. Instead of forcing everyone to prove who they are, you put the burden on sites to declare themselves safe for kids if they want that traffic. It’s a clever, engineering-first approach to a messy legal problem. And then there’s Kagi. The search for a Google alternative is, let’s be real, a permanent fixture in tech conversations. But Kagi is actually getting traction. People who use it, love it. A user who’s been on it for a year said they “couldn’t recommend it enough,” that the results are what they’re actually searching for. But here’s the catch, and it’s a big one. Right now, Kagi is still pulling from other indexes, including Google’s.

It’s a great user interface, a great layer of AI and personalization… on top of data it doesn’t fully control. A Kagi employee jumped into the thread to clarify. They said, yes, we know, and “we’re actively working on building our own indexes… not just a general index but also purpose built indexes for things like programming.” So they’re trying to solve the problem at the source. But building a search index from scratch today is a nightmare. You’re not just fighting to crawl the web, you’re fighting SEO slop and what one user called “Anubis pages”—content specifically designed to trick and evade search engine crawlers. So you have three stories. One, a debate over whether AI is good enough to write our code.

Two, a fight over how to regulate a chaotic internet without breaking it. And three, a desperate search for a tool that can find truth in an ocean of garbage. It’s all the same story. It’s all about quality, and who’s in control of it. So let’s go back to that number. Eighty to ninety-five percent. When an LLM gets you ninety-five percent of the way there on a coding task, that feels like magic. But what lives in that last five percent? Bugs. Security vulnerabilities. Edge cases. The stuff that separates a hobby project from enterprise-grade software. As one user put it, “LLMs can get you 80-95% of the way, sometimes less, sometimes more. And heck, sometimes, it just gets you somewhere wrong.” That’s the problem.

It’s not just that it’s incomplete; it’s that it can be confidently, plausibly, disastrously wrong. And this is where the conversation got really sharp. Someone posted a quote from Shakespeare, from Julius Caesar: “The fault dear Brutus isn't in our stars, but in ourselves.” And they applied it here. “The fault dear Brutus isn't in our language models, but in ourselves.” Their point was that the technology isn’t the problem. The problem is our expectations. We want a magic button. We want an autonomous robot coder. What we’re getting is a tool. A powerful, flawed, brilliant, and sometimes stupid tool. And we haven’t figured out how to use it yet. This is where we’ve seen this before.

This is not a new story. One comment absolutely nailed it. They said, “AI is the first threat to the Artisans of today (ie programmers). We are just the most recent in a long history of Luddites.” Now, hold on. The word Luddite gets thrown around as an insult. It means you’re afraid of technology, you’re anti-progress. But that’s not what the Luddites were. They weren’t afraid of machines. They were skilled textile artisans in the early 1800s, and they were protesting the specific machines that were being brought in to replace them. Why? Because those machines produced garbage. They took a craft that required years of expertise and replaced it with a machine that churned out inferior, low-quality cloth, just to save money.

They weren’t fighting the future. They were fighting the erosion of quality. They were fighting for their craft. Sound familiar? That’s the fear behind “The Eternal Sloptember.” It’s not that AI will replace programmers. It’s that it will devalue the craft of programming. That the market will be flooded with 95-percent-correct code, and that “good enough” will become the new standard. That the meticulous, detail-oriented work that makes up that last five percent—the work that prevents outages, that secures data, that makes software a pleasure to use—will be seen as an expensive, unnecessary luxury. Of course, the analogy isn’t perfect. A critic in the thread immediately pointed out that while a hand-crafted wooden table is still better than one from Ikea, a modern CNC machine produces parts with a precision no human hand could ever match.

The Luddites were protesting machines that made worse stuff. In many fields today, from semiconductors to automotive manufacturing, the machines are objectively better. So where does software fit? Is it more like a hand-carved table, where the human touch and intuition for the material is everything? Or is it more like a semiconductor, where machine precision is the only way to achieve the required quality? The answer is… it’s both. And that’s what makes this so complicated. A lot of coding is boilerplate. It’s plumbing. It’s connecting API A to database B. It’s work a machine can and probably should do. That’s the CNC machine part. But designing the architecture, understanding the business logic, anticipating user behavior, securing the system against novel threats… that’s the artisan part.

That’s the part that requires taste, and judgment, and a deep understanding of the medium. The fear isn’t that the machine will take over the whole process. The fear is that we’ll get so excited about automating the 95 percent that we forget the last five percent is the only part that really matters. We’ll fire the master carpenters and then wonder why all our Ikea tables keep collapsing. This is why the Kagi story is so relevant. The reason people are flocking to a paid search engine is because the quality of the free one, Google, has degraded. It’s filled with SEO spam, AI-generated content, and ads. It’s a system that has been optimized for engagement and revenue, not for delivering the best answer.

It’s a system that has prioritized “good enough” for the masses over excellence for the user. People are now willing to pay money just to escape the slop. They are paying for a curator. For a tool that respects their time and values quality. And look at the California law. The government tried to apply a blunt, one-size-fits-all solution to a complex ecosystem. And the tech community’s response wasn’t just “no.” It was, “here is a better, more elegant, more precise solution.” The RTA header proposal is an act of craftsmanship. It’s an attempt to solve the problem with the grain of the internet, not against it. So what this week is really showing us is a pattern.

In every corner of the tech world, we’re seeing a battle between mass-produced, 95-percent-good-enough solutions and the demand for true quality, for craft, for the last five percent. The debate isn't really about AI or search engines or regulation. As one user put it, framing the whole thing in a much starker light: "It's not 'competition'. It's carnivorous, predatory. Consider shifting gears and seeing all of this through the lens of 'power'." What we're seeing this week is a fight for who holds that power. The power to define what "good" means. Is it the platform that can generate the most content the fastest? Or is it the artisan who knows how to get it right?

This week sets up a conflict that will define the next decade of technology. This isn't a simple story of new tools replacing old jobs. This is a fight for the soul of the craft.