Why agents matter more than other AI
Josh Albrecht argues that among all AI categories, agents — defined as systems that "run tools in a loop to achieve a goal" — are the only type that truly matters for the future, because they remove humans from the loop entirely. Other AI forms (chatbots, image generation, recommendations) are constrained by limited human attention, but agents are limited only by compute spend.
The Coding Agent Precedent
The author points to coding agents as proof of concept. His own company, Sculptor, builds such tools. Albrecht notes that a non-programming designer on his team used these tools to "make a full clone of 'Geometry Wars' that is honestly better than some of the sequels." Cursor and Lovable have seen revenues skyrocket, with some users spending over $1K per day.
Coding agents succeed because they are safe — bad code can simply be deleted, costing only pennies. Other early-fit domains include research reports, presentations, drafting emails, editing documents, and filtering information.
Seven Advantages Over Human Employees
-
Infinite replication — Improvements to one agent instantly propagate to all copies.
-
24/7 operation — Agents don't need rest, sleep, or food.
-
Faster throughput — They can write code (and eventually think) faster than humans.
-
No management overhead — No performance reviews, growth plans, or emotional needs. "Human managers get some level of importance and status from having human reports," slowing adoption.
-
Instant scaling up and down — Agents can be deployed en masse and shut down immediately. "Humans...get pretty annoyed when you stop paying them."
-
No privacy concerns — "Agents don't mind having their every action watched in excruciating detail." This reduces security risks like fraud or backdoors.
-
Tax efficiency — Agent costs are treated as business expenses or R&D, avoiding payroll taxes, Social Security, and healthcare costs.
Employment Impact
The author cites a Bloomberry analysis showing front-end developer job openings down nearly 10% year-over-year, and professions like photographer and writer down nearly 30%. He sees a near future where companies "hire" agents instead of employees, describing a colleague who let a coding agent run during lunch while playing chess.
Closing Questions
- "Will these agents be working for you, or will they be maximizing profit for some company?"
- "What kind of future are we building as the value of human mental labor approaches zero?"
- "What should we be doing today to prepare for (or prevent) this future?"
He references a tweet by Ethan Mollick noting the latest models have surpassed human performance on GDPval, a benchmark of economically valuable knowledge work tasks.
Internal Link
The author refers to a related piece on whether AI agents will work and cites Simon Willison's definition of agents.