Maximalism of the Gaps

0001-01-01

An LLM "maximalist" is someone who has faith that LLMs have an effectively infinite potential - that they’ll reshape the species itself, and possibly even become AGI and a threat to humanity. At the very least, it’ll cut the number of software engineering jobs, since good engineers will become 10x more effective and we simply don’t need as many of them.

We haven’t seen any of this bear out, though. Widespread tech layoffs are happening, but [[De-Investment Causes Layoffs, not AI]]. It can’t become AGI because [[Models Have a Limit]].

A maximalist will always argue that there is an LLM solution to LLM problems. If the model isn’t producing the right code, or it’s misdiagnosing problems, or not following links that it ought to, then that means you have to write more subagents/skills/plugins to explicitly tell it not to do that; running afoul of [[Userland Is Not Enough]].

Sometimes, you can carbon-date when someone started using LLMs based on their position. Maximimalists often have only used Claude, haven’t shipped anything to production with it, and spend most of their time making agents, skills, plugins, and tools to work around LLMs. There are very few maximalists who actually run production systems, most of the time they’re just having fun making a little agent ecosystem.