Hello! Another huge batch of links for you this morning. I really need to work on more frequent, smaller updates—it's been almost a month since we last talked.

How are you, friend? Are you sick of AI yet? Or just sick of the LinkedIn posts about it? As you'll see from several of the posts below, I've been working on swinging the pendulum back from "let the AI cook on everything" to a more thoughtful figuring out of when artisanal hand-crafted words are the way to go. It's not rocket science of course, but helping with minutia/summarizing = good. Outsourcing thinking = bad. Thanks for coming to my Ted Talk.

Have a good week! Rian

Link Posts

• I am dreading our LLM-written incident report future — Lorin Hochstein on why AI-generated incident reports are riskier than AI-generated code: unlike code, a write-up has no obvious test for correctness, so plausible errors sail straight through.

• 'What I see in clinic is never a set of labels': are we in danger of overdiagnosing mental illness? — GP Gavin Francis in the Guardian asks whether broadening diagnostic criteria is starting to pathologize ordinary distress, and whether understanding difficult emotions beats trying to label them away.

• The Cult of the Enhanced Self — Derek Thompson on how fear of mortality fuels health optimization that crowds out the social connections research says matter as much as exercise for a long life.

• Do Not Resign From Life — L.M. Sacasas argues that AI being good at thinking and writing is no reason to stop doing either, because the meaning lives in the doing.

• Guidelines for Respectful Use of AI — Camille Fournier's rule that every team should adopt: don't ask a colleague to review AI output you haven't read yourself.

• We Should Be More Tired Than the Model — Vicki Boykis makes the case for deliberately slowing down with AI tools so you actually understand what you ship instead of optimizing for raw speed.

• The Great AI Cost Panic of 2026 — Derek Thompson reframes the hand-wringing over AI token spend as the normal under-then-over-investment cycle every technology goes through.

• Your AI Use Is Breaking My Brain — A 404 Media piece on catching yourself writing in telltale AI patterns even when you never touched a chatbot, and what that saturation does to how we all write.

• LLMs and Buttondown — Justin Duke on how Buttondown's API-first design made it the default newsletter platform for LLM developers, nearly doubling growth month over month.

Side project releases

• Discrobble v1.3.1 — Moves to a freemium model (20 free scrobbles, then a $4.99 unlock) and adds a Wikipedia summary on each album; early users keep unlimited free access.

• Discrobble v1.2.1 — Adds album ratings: see your Discogs rating and rate albums 1-5 stars in the app, synced back to Discogs.

• Discrobble v1.2.0 — Syncs your Recently Scrobbled history across devices via Last.fm, with history clearing that stays in sync everywhere.

• Discrobble v1.1.0 — Native iPad support with adaptive layouts and side-by-side album/tracklist views, plus AES-256 encryption for auth tokens and full data deletion on account close.