Cyborgs Will Kill the Corporation

Octopusyarn argues that AI agents aren't assistants bolted onto humans — they're an exoskeleton that turns the human+agent pair into a single economic entity (a cyborg). When every knowledge worker becomes a cyborg, the corporation — invented to reduce transaction costs between humans — becomes a liability rather than an advantage. The piece forecasts firms decomposing into three forms: excorporations (temporary collaborations between cyborgs, dissolving when the work is done), plankton (fully automated commodity services), and protocols (infrastructure like Stripe or Uniswap that shed their institutional wrappers). The harder argument: judgment, taste, and trust remain irreducibly human because they're grounded in embodied cognition and finitude.

Key Quotes

"The corporation is a machine for reducing the cost of getting humans to work together."

This is the Coasean hinge of the entire argument. If the corporation exists only because coordination is expensive, and AI makes external coordination cheap, the firm's reason for being evaporates.

"good judgment looks more like vibes than algorithms"

A direct hit on the idea that AGI replaces human decision-making. Judgment is embodied, experiential, and resists formalization — which is why the cyborg model keeps a human in the loop, not as a rubber stamp but as the irreplaceable core.

"Skin in the game requires skin, attached to a nervous system that feels consequences."

The trust argument in one sentence. Two cyborgs collaborating is still two humans choosing to be vulnerable. The agent exoskeleton doesn't change the calculus of who bears the risk.

"Intelligence is a commodity now, rentable by the token."

Dismissed almost in passing, but it's load-bearing. If intelligence is cheap and abundant, what's scarce shifts to taste, judgment, and relationships — the things the human interior provides.

"Picking low-hanging fruit is for plankton."

The closing provocation: competing on efficiency makes you interchangeable. The only durable play is the relationship layer — the "soft part" inside the cyborg that agents can't replicate.

Key Themes

• #concept Excorporation — The inversion of incorporation: collaboration constituted at the intersection of cyborg exoskeletons, existing only while interaction continues. No bylaws, no persistence, no legal body.
• #concept Cyborg as economic unit — Human interior (judgment, taste, trust) + AI exoskeleton (execution, coordination, automation). The cyborg molts agents as goals change.
• #pattern Corporate decomposition — Firms unbundle into excorporations (judgment-heavy work), plankton (automated commodity services), and protocols (infrastructure without institutional wrappers).
• #concept Embodied cognition as moat — Intuition, taste, and trust live in the nervous system and can't be uploaded. This is what keeps humans essential even when intelligence is token-rentable.
• #concept Rhizomatic trust — Multilateral, web-like permissions replacing hierarchical authority. Trust as infrastructure rather than a byproduct of employment contracts.

Critical Analysis

The essay's greatest strength is its refusal to be a tech-optimist or tech-pessimist piece. It's an economic theory essay dressed in cyborg metaphors, and it works because Coase gives it real analytical scaffolding. The transaction-cost collapse argument is the same one underpinning AI Killing B2B SaaS, The Mythical Agent-Month, and Zero Alignment — but Octopusyarn takes it further, predicting not just corporate efficiency loss but corporate form loss.

The cyborg framing solves a tension that runs through much of the wiki's agent literature. Agent Identity argues agents need identity grounded in participation; Smart Models Dumb Pipes separates judgment from execution. Octopusyarn's cyborg model does both: identity lives in the human interior, execution in the exoskeleton, and the boundary is deliberately blurry.

The weakest link is the trust argument. "Skin in the game requires skin" is rhetorically satisfying but dodges the hard question: what happens when agent exoskeletons become good enough at simulating trustworthiness that humans treat them as trusted anyway? The essay waves at this with "trust requires bodies" but doesn't engage with the possibility that humans are demonstrably willing to trust non-human systems (markets, courts, code) when they produce reliable outcomes. Cybersecurity Is Proof of Work Now and The Future of Everything is Lies I Guess both offer harder-edged takes on what happens when the simulacrum becomes good enough.

The three-destinations taxonomy (excorporations, plankton, protocols) is clean but undertheorized. The "plankton" category especially feels hand-waved — "automated processes that drift through the economy as commodity services" is a description, not an analysis of who owns them, who maintains them, or what happens when they fail. Long Live Systems of Record offers a concrete counterpoint: someone still needs to answer "where does the truth live."

That said, the essay lands a genuinely novel concept in "excorporation." It's not just a neologism — it names something the wiki's agent orchestration pages keep circling without quite articulating: the possibility that the durable organization is the wrong unit of analysis entirely. Scaling Long-Running Agents, Agent Orchestration, and Zero Alignment all grapple with coordination across agent teams, but they still assume coordination happens within some persistent entity. Excorporations suggest the entity itself is temporary — and that's the point.

Sources: Cyborgs Will Kill the Corporation — On Excorporations and the Death of the Firm Last updated: 2026-05-15