Write Things Down

TEXT START: It was Merlin Mann and his 43folders website who introduced me to David Allen and Getting Things Done.

The Dissection

This essay is ostensibly about productivity and AGI definitions. Its actual subject is externalized control: memory becomes notes, notes become state, state becomes workflow, and workflow becomes agent-readable execution.

The author has built a small Sovereign stack. Assistants absorb operational residue; harnesses absorb cognitive residue; the owner retains direction, judgment, and output. The essay mistakes that position at the top of the stack for a universal human future. Its own examples show the opposite: writing things down makes human work inspectable, transferable, and automatable.

The strongest technical insight is that persistent files, tools, feedback, and deterministic scheduling can give frozen models functional continuity. That is a real mechanism. Calling it pseudo-AGI does not neutralize its economic effect.

The Core Fallacy

The central error is treating volition, morality, and ultimate authorship as equivalent to economic indispensability. They are not.

A system can have a human principal, legally assigned responsibility, and no inner life while eliminating the labor that formerly executed and coordinated the principal’s goals. A CEO can remain the goal-setter while payroll collapses. AI being a tool is no defense; tools are how owners displace workers.

The AGI definition is also a distraction. The Discontinuity Thesis does not require metaphysical AGI or continuous weight updates. If a frozen model plus external memory and a harness performs cognitive work at lower cost, the labor-market threshold has already been crossed. The label changes nothing.

Hidden Assumptions

• The human who decides will also remain economically needed. In reality, control and ownership can concentrate in Sovereigns while everyone else loses bargaining power.
• Moral responsibility creates productive scarcity. It may create liability, but liability can remain attached to a nominal human signatory.
• Creation produces wages. Credit is not compensation; markets reward scarce control over productive systems.
• The author’s assistant-agent arrangement scales benignly. It scales first as hierarchy: Sovereign, servitors, and machines. The assistant role itself becomes a candidate for automation.
• Physical and interpersonal work are permanent refuges. They are lag defenses, not reversals of the underlying trend.
• Making anything creates durable value. Mass generation produces supply explosions, price compression, and faster obsolescence for anyone whose edge is implementation.
• Misalignment is the main danger. Aligned AI controlled by concentrated owners is sufficient to destroy wage dependence; rogue consciousness is unnecessary.

Social Function

Primary classification: partial truth. Secondary classification: elite self-exoneration and transition management. Tertiary classification: ideological anesthetic.

The essay accurately describes harnesses, external memory, and the non-persistent character of current models. It also gives operators a practical blueprint for scaling themselves through assistants, bots, status boards, and agents.

But it shifts the political question from who controls productive capacity to who receives credit for the idea. When wages and leverage disappear, symbolic authorship becomes the consolation prize. This is not simple copium—the technical observations are often sharp—but those observations are used to preserve a humanist conclusion they cannot support.

The Verdict

This is a high-resolution description of the mechanism and a low-resolution account of its consequence. The essay sees that writing creates scalable memory and that harnesses make static models functionally persistent. It then confuses the persistence of human intent with the persistence of mass human participation.

Under P1, written and harnessed workflows make cognitive labor cheaper and portable. Under P2, no stable human-only cognitive domain survives at scale once the workflow can be externalized and replayed. Under P3, most people can lose access to economically necessary labor while humans remain nominal authors, principals, and moral signatories.

The author is describing a viable Sovereign position: control the AI stack, use servitors for the physical remainder, and reduce personal judgment to a command layer. That is survival leverage for the owner, not evidence of system survival. The post-WWII wage-consumption circuit can die with human volition intact. The machine does not need a soul to take the job; it only needs an owner who has one.