Cognitive Scaffolding

An environment deliberately structured so it carries part of the cognitive load: post-it notes in the right places, ingrained routines, webs of notes and off-the-shelf software. The source's two anchor cases: inner-city Alzheimer's patients whose test scores said they could not live alone yet who functioned in homes structured with notes and routines — their minds "kind of distributed across their environments" — and Patrick Jones, a deacon with Memento-style traumatic brain injury who holds down his job through "huge webs of Evernote structures" and consults his iPhone to recall a conversation he just left. The defining consequence cuts both ways: the scaffold raises capability beyond what the unaided person tests at, and removing it degrades the person like an injury — relocating the patients to care homes was "as if someone had reached into your brain overnight and lesioned it a bit."

For the vault: an agent's harness — its CLAUDE.md routing, memory files, transcript store — is scaffolding in exactly this sense, so capability is a property of the agent+scaffold system, and the scaffold's integrity is a first-class concern, not a tooling nicety.

Claims

Capability lives in the person+scaffold system: assess in the structured environment, not in isolation — the tests said the patients could not function; in situ they could. (principle — as asserted by Clark via his colleague's observations)
Removing or relocating someone away from their scaffold degrades functioning like a brain lesion — interference with the environment can be interference with the person. (principle — as asserted by Clark; the ethical edge of Extended Mind)
The video recounts that structured homes (post-its, routines) let Alzheimer's patients live alone against their test predictions, per Clark's occupational-therapist colleague at Washington University in St. Louis. (observation — second-hand anecdote; check-worthy)
The video states Patrick Jones, a Catholic deacon in Colorado Springs with traumatic brain injury, functions via Evernote webs and off-the-shelf software. (observation — the source's example; check-worthy)
"Test the person with the phone, not the person without the phone — that person is irrelevant": assess the whole extended person, scaffold included. (best practice — context: testing/education where the tool is a stable part of the person's future working configuration; as asserted by Chalmers in the sibling interview. The agent-eval reading: benchmarking an agent stripped of its harness/memory measures a unit that will never ship)
Operating with an AI-powered second brain will change the biological brain: the human's skills shift toward probing questions, pattern identification, and connection-drawing — the person co-adapts to the technological appendage. (observation — the 2023 Contrary Research report's forward-looking claim; the co-adaptation direction of the person+scaffold system)
A source contradicts the "test the person with the phone" claim for generative scaffolds: it holds that judgment is the one function whose offloading is a net loss, so the unaided person is not irrelevant but load-bearing. (principle — as asserted by a contradicting source; unresolved) — every scaffold on this page holds what the person already judged (post-its, Evernote webs, routines). A model that writes the essay judges on the person's behalf, and Sandeep Swadia argues that in an era when the scaffold can fabricate (Authenticity Collapse), someone outside the scaffold has to adjudicate it — an adjudicator this page's own logic would let atrophy. His evidence is an MIT essay study he says found ChatGPT users showed the least brain activity and could not quote a line of their own writing minutes later (second-hand and check-worthy). The tension is real and unresolved: this page says assess the extended person, scaffold included; that source says the unaided person is the whole ballgame. The reconciliation worth testing is that holding is safely offloadable while judging is not, in which case both survive because they are about different functions. Nothing in either source settles it, and no vault source has yet run this page's scaffold-removal test on a generative scaffold. See Cognitive Offload Cost, which carries the open question.

Extended Mind — the thesis; scaffolding is what the ethical route runs on.
Cognitive Offload Cost — the contradicting position: offloading has a price paid in the person, and a generative scaffold is the case the parity move was not formulated against.
Parity Principle — the test that admits a scaffold into the cognitive system.
AI Second Brain — the operator-built scaffold: files and routing that raise the operator+agent system's capability.
Layered Agent Memory — an agent's own scaffold, tiered by freshness and cost.
Distillate: Andy Clark — What is Extended Mind? (Closer To Truth)
Distillate: David Chalmers — What is Extended Mind? (Closer To Truth)
Distillate: From Notetaking to Neuralink (Contrary Research) — the co-adaptation claim: the scaffold reshapes the person as well as extending them.
Distillate: This Skill Makes You Dangerous In The AI Era — the contradicting source: judgment as the function that must not be offloaded, and a generative scaffold as the case that breaks parity.

Cognitive Scaffolding

Claims

Related

Linked from