No Sloptober: a month-long challenge to work without LLM tools

The website no-sloptober.com proposes an experiment for October: spend the month without LLM-based tools. The page calls it “a fast for your mind” and stresses that it is a personal challenge, not a judgement of other people’s workflows.

The rules of the fast

The suggested abstention covers tools many developers now touch every day: AI summaries in search results, chatbot chats and chatbot-assisted search, AI code review, and “Claude/OpenCode/Pi/Agents/Models etc.” at home or at work. The advice is to “do it the hard (core) way” and to work out your own sense of what LLMs are good and bad at. Two footnotes soften the rules: employees under corporate pressure to use AI are told to adjust the challenge to keep their jobs, and machine translation is offered as a personal exception — although the author argues that professional translation and localization carry an accuracy requirement that should not be delegated to a machine.

Why the author asks developers to try it

Skill upkeep comes first: even people who find LLMs useful should know how to do the work themselves. The page also points to perspective and mental health, and links Simon Willison’s essay “Don’t be a meat proxy” of 3 August 2026, which describes people who launder conversations through a chatbot and apply little or no editorial review to the output. Security is a further motive, with a parenthetical nod to agents that install malware by accident. The page also prints a line it labels Onarheim’s Law: “Agents can only maintain or INCREASE entropy in a system. Humans are uniquely capable of decreasing it”.

What to do instead

For participants, the site suggests concrete replacements: review and clean up existing LLM output in work and home codebases, which “really do like to write A LOT of cruft in their comments” and over-engineer software; learn a language such as Zig, Rust, Go, Lua or TypeScript; pick up a web framework like Astro, Mastro or Effect.ts; practise code katas; start a project; or build a game with Excalibur or Godot. Teams get a suggestion too — run a cost and risk analysis on LLM usage and measure velocity, incident rate and spending under reduced use, remembering the “iron triangle”: good, cheap or fast, pick two.

What the month is meant to show

The closing argument is about craft. The page lists finding joy in the work again, returning to flow, and asking which tasks have no value and could be automated deterministically instead. Learning, it notes, takes mental effort and friction. Participants are invited to write about the experience and post it with the hashtag #no-sloptober.