Les Orchard has been building things on the internet for a long time. A software developer, blogger, and self-described tinkerer who has worked at Mozilla and contributed to the open web for decades, Orchard is the kind of person whose opinions carry weight among the engineers and writers who shaped the internet before it became a platform for trillion-dollar companies. So when he published a deeply personal blog post in March 2025 titled “Grief and the AI Split,” it didn’t just land as one more take in the AI discourse. It landed like a eulogy.
“I’m grieving,” Orchard wrote on his blog. Not over a person, but over relationships — friendships and professional connections strained or severed by fundamental disagreements about artificial intelligence. The post describes a growing rift among technologists, one that cuts through friend groups, open-source communities, and the social fabric of the very people who built the tools we all use. It’s a split that doesn’t map neatly onto left versus right or corporate versus indie. It runs deeper than that.
The argument, as Orchard frames it, isn’t simply about whether large language models are useful. It’s about consent, labor, environmental cost, and what happens when the tech industry vacuums up the creative output of millions of people to train systems that may ultimately replace them. He describes watching friends he’s known for years become enthusiastic adopters — or even builders — of AI products, while he and others feel their work has been extracted without permission. The grief is real. And it’s spreading.
This isn’t an isolated sentiment. Across the open-source world, among independent web developers, digital artists, writers, and the broader creative-technical class, a fault line has opened. On one side: those who see generative AI as a natural, even inevitable extension of computing’s trajectory. On the other: those who view it as an unprecedented act of appropriation, one that treats the collective output of human creativity as raw material for corporate profit.
What makes Orchard’s account so striking is its intimacy. He’s not writing a policy paper. He’s talking about people he likes — people he’s shared meals with, collaborated with, argued with in good faith for years. The AI question has introduced a moral dimension that makes continued collegiality feel, to him, untenable. “It’s like finding out a friend works for a company you find ethically repugnant,” he writes. “Except worse, because they’re excited about it.”
The comparison is imperfect, and Orchard knows it. But the emotional texture is precise.
To understand why this split has become so bitter, you need to understand what the open web meant — and still means — to the people who built it. For a generation of developers who came of age in the late 1990s and early 2000s, the web was a shared commons. You published your code. You shared your writing. You contributed to open-source projects and Creative Commons repositories because you believed in a set of principles: that knowledge should be free, that collaboration beats competition, and that the network’s value comes from the people who populate it. The ethos was reciprocal. You gave, and others gave back.
Generative AI broke that contract. Or at least, that’s how a significant portion of this community sees it. Companies like OpenAI, Google, Meta, and Anthropic trained their models on vast corpora of text and images scraped from the open web — much of it created by the very people now objecting. The training data included blog posts, forum threads, Stack Overflow answers, GitHub repositories, Flickr photos, DeviantArt illustrations, and millions of other artifacts of human effort. In most cases, no one asked permission. In many cases, the creators weren’t even informed.
The legal questions remain unresolved. Multiple lawsuits are working through U.S. courts, including high-profile cases brought by The New York Times against OpenAI and by visual artists against Stability AI and Midjourney. But the legal arguments, however important, are almost secondary to the emotional and philosophical breach that Orchard describes. The law may eventually say this was fair use. That won’t make the people whose work was consumed feel any less violated.
And the rift isn’t just philosophical. It has practical consequences. Open-source maintainers are changing licenses to explicitly prohibit AI training on their code. Artists are poisoning their images with tools like Glaze and Nightshade to disrupt model training. Writers are adding robots.txt directives and AI-specific blocks to their sites. Some are leaving platforms entirely — abandoning GitHub, pulling work from public repositories, going dark. The commons is contracting.
Orchard’s post captures something else, too: the social awkwardness of the split. In a community that prizes technical rationality, accusing someone of ethical failure feels transgressive. The AI enthusiasts often frame their position in terms of progress, efficiency, democratization of tools. The skeptics counter with arguments about exploitation, consent, and environmental harm. Neither side thinks it’s being unreasonable. Both sides think the other is missing something fundamental.
This dynamic has played out with particular intensity on Mastodon and the broader fediverse, the decentralized social media platforms that became a refuge for many open-web advocates after Elon Musk’s acquisition of Twitter. On these platforms, the anti-AI position is something close to a consensus norm. Posting enthusiastically about using ChatGPT or generating AI art can get you unfollowed, muted, or blocked. Server administrators have written policies banning AI-generated content. The social cost of being pro-AI in these spaces is real and immediate.
But the fediverse is not the whole internet. On LinkedIn, on X, on Hacker News, and in corporate engineering departments, AI adoption is accelerating. GitHub Copilot has millions of users. Companies are integrating LLM-powered features into every product they ship. Venture capital continues to pour into AI startups at a staggering rate — Reuters has reported that global AI investment exceeded $100 billion in 2024 alone. The market has made its bet. And many of the people Orchard once considered allies are now on the other side of it.
The grief Orchard describes is, in part, the grief of watching a shared value system dissolve. The open web was never a monolith, but it had a center of gravity: a belief that technology should serve people, not extract from them. That center is gone. Or rather, it’s been claimed by both sides. AI boosters say they’re democratizing access to powerful tools. AI skeptics say they’re watching the largest transfer of creative value in history, executed without compensation or consent.
There’s a generational dimension to this, though it’s easy to overstate. Younger developers who grew up with GitHub Copilot and ChatGPT often see these tools as simply part of the stack — no more morally fraught than using a compiler or a search engine. Older developers and creators, who remember a web before platform monopolies, tend to feel the loss more acutely. They built the commons. They watched it get enclosed. And now they’re watching the enclosure get automated.
But age isn’t destiny here. Plenty of younger technologists are deeply opposed to AI training practices, and plenty of veterans have embraced the tools. The split doesn’t follow demographic lines so much as it follows lines of experience and exposure. If your livelihood depends on writing, illustration, or other creative work that LLMs can approximate, you’re more likely to see the threat. If you’re a backend engineer whose productivity jumps 30% with Copilot, you’re more likely to see the benefit.
Orchard doesn’t claim to have answers. His post is explicitly framed as an expression of loss, not a manifesto. But its resonance — it was widely shared and discussed across the fediverse and independent web — suggests he’s articulating something many people feel but struggle to say. The tech industry has always had internal disagreements. What’s different now is the moral weight. This isn’t a debate about tabs versus spaces or REST versus GraphQL. It’s a debate about whether the foundational act of the current AI boom — scraping and training on human-created content without consent — is acceptable. And that’s not a question you can resolve with a pull request.
The broader industry has largely ignored this internal fracture. Major tech companies continue to ship AI features at breakneck speed. Microsoft, Google, Apple, Amazon, and Meta are all racing to embed generative AI into their core products. The financial incentives are overwhelming. Wall Street rewards AI investment; it punishes hesitation. In this environment, the concerns of open-web veterans and independent creators register as background noise — sincere, perhaps, but commercially irrelevant.
That dismissal may be premature. The legal cases could reshape the training data question. Regulatory frameworks are emerging in the EU, with the AI Act imposing new transparency requirements on model training. In the U.S., the Copyright Office has been studying the question of AI and intellectual property, and congressional hearings have featured testimony from artists and writers whose work was used without permission. The policy environment is shifting, even if slowly.
And there’s a cultural dimension that’s harder to quantify but no less real. Trust matters in open-source communities. Collaboration depends on it. If a significant number of contributors feel that their work is being exploited, they’ll stop contributing. They’ll lock down their code, restrict their licenses, and withdraw from the commons. That’s already happening. The long-term consequences for open-source software — which underpins virtually every major technology product in existence — could be substantial.
Orchard’s post ends not with resolution but with an honest admission of uncertainty. He doesn’t know how to maintain friendships across this divide. He doesn’t know if the divide is permanent. He knows only that something has broken, and he’s mourning it.
That mourning deserves to be taken seriously. Not because the anti-AI position is automatically correct, but because the people expressing it aren’t luddites or cranks. They’re the builders. They wrote the code, designed the protocols, authored the content, and maintained the infrastructure that made the modern internet possible. When they say something has gone wrong, the industry would be wise to listen.
Whether it will is another question entirely.
WebProNews is an iEntry Publication