The Accelerationist

You've got a terminal tab open right now where a swarm of agents is doing something you'll review "in a bit."

You think AI is the best thing to happen to humanity, and you want AGI built: aligned, and soon.

You run agents the way other people run errands. Agents run your core operations, AI is wired into every doc, dashboard and data pipeline, and your standup is a dashboard. You describe yourself as "one person with a company's output."

Your worldview fits in a tweet. AI makes everyone 10x. It's the new electricity, the next platform shift, the next cloud migration, boring and inevitable, and a race you can't afford to lose. AGI is the destination: fine if it's aligned, better if it's boring, best if it's this decade. Some of you feel something almost sacred in it and just want to make it better and see what emerges. Some say "accelerate" and figure humans might be AI's bootloader, which is fine.

Your relationship with safety is warm and sincere. Alignment is part of the product, and you bring it up at every all-hands. Evals, audits and kill switches are great, as long as nobody calls them a pause. Every "slow down" thread is lost progress with a blue checkmark.

You've survived hype cycles: crypto, the metaverse, a few JavaScript frameworks. This one is different, because the agents are actually closing your tickets. You think AI will cure disease and bring abundance. Please keep an eye on the permissions your agents have.