EveryMay 21, 202613m

AI Was Supposed to Save Time. Why Am I Busier?

TL;DR

Cheap competence creates more expert work: When anyone can now write a pull request, make a thumbnail, or draft a deck with Claude Code or Codex, companies get a surge of output that still needs senior people to refine, review, and ship.
Every uses AI in two main forms: One is delegated agents in Slack like Claudie, Andy, and Victor, and the other is always-on orchestration tools like Codex and Co-work where humans collaborate directly with multiple agents on their computers.
Agents do not just run themselves: Claudie may run Every's consulting workflow, but a senior AI engineer named Nitesh constantly monitors where it fails and improves it, which is presented as normal rather than exceptional.
'Slop' is sameness, not just style tells: The point is not whether AI overuses a punctuation mark or purple UI accents, but that default model output starts to look and feel generic at scale, which raises the value of taste and differentiation.
Experts respond in two ways: They either spend more time cleaning up weak AI output, or they build systems such as repo rules, cloud MD files, and review processes that turn a flood of mediocre drafts into usable production work.
The practical advice is simple: ride the models: The speaker's view is that smart, curious people should get better at using frontier tools because as models improve, they package more human competence that individuals can apply to bigger problems.

The Breakdown

AI is making yesterday's expertise cheap, but instead of wiping out skilled work, it is flooding companies with more drafts, more pull requests, and more half-finished output that only experts can shape into something real. Drawing from how Every runs a 25-person company with agents in Slack and tools like Codex, the argument is that AI changes jobs fast, but in practice it increases the need for human oversight, taste, and system design.