---
title: Outcomputed
description: Jan Wilmake's concept describing the inevitable point at which AI surpasses a person's professional utility, and his proposed response — becoming a compute allocator
category: Concepts
---

# Outcomputed

**Outcomputed** is a concept introduced by [[Jan_Wilmake]] in a 2024 essay describing the trajectory of human labour utility as artificial intelligence capabilities compound. The core claim: for any given knowledge worker, there is a point in time — their personal "outcomputed" threshold — at which machines can perform their primary work tasks better than they can. Once crossed, the conventional basis for that person's economic participation largely evaporates.

Jan's own framing: *"GPT-4o is reported to be smarter than highly educated knowledge workers (knowledge-wise) and this trend is continuing. The % of tasks that computers can do will steadily go up over the coming years. In any case, there will be a point in time at which you'll be outcompeted by machines: Outcomputed."*

## The Human Utility Curve

Jan sketched a "human utility curve" showing how professional value will trend over time:

- **Already underway**: certain niche professions where AI has crossed capability thresholds (writing, image generation, some coding tasks)
- **Near-term**: a rapid general collapse once AI crosses a threshold of *general* intelligence, affecting the largest segment of knowledge workers simultaneously
- **Survivors**: people working at the technical frontier and fast adaptors; the curve stays elevated for this group

The speed of collapse, Jan noted, depends heavily on regulatory response and how quickly new AI capabilities reach production scale — factors that are difficult to forecast.

## The Compute Allocator Response

Jan's proposed strategy for staying relevant draws on an insight from Dan Shipper (Every.to): rather than competing with AI at the object level, become a **compute allocator** — someone who directs, evaluates, and composes AI systems to produce outcomes.

The analogy: in the current world, leverage comes from directing other humans to work for you (management, delegation, hiring). In the AI era, leverage comes from directing machines. The fundamental skill shifts from domain expertise to **meta-level orchestration**: knowing which AI systems to deploy, how to evaluate their outputs, how to chain them, and how to identify which problems are worth solving.

Jan's own bet: *"My bet is to build highly reliable agents, so I can enable this compute allocation to be highly effective in the real world — both for myself and for others."*

This directly informed his decision to invest in agent infrastructure ([[uithub]], [[sponsorflare]], [[ActionSchema]]) rather than building end-user AI applications that would themselves be outcompeted.

## Connection to Other Frameworks

The Outcomputed concept is a specific application of the [[The_World_Is_Information]] framework: if all phenomena are information streams, the valuable human skill is processing those streams to generate optimal actions — i.e., allocating compute toward the highest-expected-value interventions. A person who can do this well becomes more valuable as AI capability increases, not less, because they can leverage exponentially more compute.

It also connects to Jan's critique of startup culture ([[Reciprocal_Networks]]): if AI can do most knowledge work, the justification for large co-founder teams collapses. One person with good compute allocation skills and a reliable agent network can build what previously required a team.

## See Also

- [[The_World_Is_Information]]
- [[Jan_Wilmake]]
- [[ActionSchema]]
- [[Scalable_Knowledge_Work]]