This batch clusters around one theme from several angles: how developers are reacting to AI, and what those reactions reveal, from community hostility to the real AI skills gap.
Armin Ronacher’s Communities of Not is a sharp short essay on a pattern he’s noticed: communities that form around abstinence from something (childfree spaces, anti-car spaces, LLM-skeptical developer spaces) tend to derive their identity from opposition. At their best they’re about something positive, but the thing being refused often becomes the main subject, and criticism curdles into policing and mob behavior when a perceived member “betrays” the group by trying the forbidden thing. He wrote it in direct response to the mob that descended on the rsync maintainer. His advice: breathe, de-escalate, default to being open to new things.
Which leads directly to the source: Andrew Tridgell’s rsync and outrage. Tridge, the original author of rsync and Samba, a developer with 40 years of experience and a CS PhD that included neural-network work, came out of blogging retirement to answer a flood of rage over his use of AI tools while hardening rsync against a wave of (often AI-generated) security reports. His account is measured and devastating: he designed the test-suite rewrite himself, used Claude/Codex/Gemini for the grunt work, reviewed every part, and ran extensive CI. His line “the world of software engineering has changed dramatically in the last few months” and his gentle roast of the openrsync crowd (failing 85 of 98 tests) are both worth reading. A primary document of this moment.
madewithlove’s The AI Skills Gap reframes the anxiety productively. After a hiring pause, Marios Prifti interviewed 40+ senior engineers from major banks and consultancies: strong backgrounds, respected companies, but little real AI fluency. No RAG, no agent frameworks, just buzzwords. His conclusion: it’s not about skill, it’s about exposure. A frontier emerged in ~18 months separating builders experimenting with AI-infused stacks from professionals following traditional workflows. Both groups are smart; only one is moving with the current. His proposed hiring signal for this era: not years of experience, but evidence of curiosity.
David Wilson’s The Solution Might Be Cancelling My AI Subscription is the most honest self-audit in this batch. He lists ~15 impressive things he built with AI (a Rust speech recognizer, a Windows 95 Notepad clone, a machine-vision traffic counter), then admits almost none are useful and he doesn’t want to maintain any. His sharpest point: this technology is “a thermonuclear ADHD amplifier,” and he’s seen the same scattered, low-commitment behavior in every adult friend using it. His frame of “friction = focus, focus = product” and his conclusion that curtailing use may be the only way to manage it land hard. Pairs naturally with the “Talk is Cheap” data piece in the next list.
Jim Nielsen’s An Ode to the Exacting Pedantry of Computers is a quiet, lovely essay. Computers have always required exacting precision: they do exactly what you tell them, and a bug is usually a flaw in your own conception, a failure to think through a case. That pedantic friction was secretly a tool for sharpening your thinking. LLMs remove it: you can have a vague thought, and the model fills in all the details you didn’t specify or anticipate. His worry, as a software user: he’d rather makers spend more time thinking so he has a better experience, than get more half-conceived stuff faster. “Intent is shaped through the act of articulating it.”