What 81,000 People Want from AI

Transparency Label: Human-only (framework)

I asked Peka to read my Reflecting on Purpose note and draw connections with Carlo Iacono's article What do you want from AI?, published the same day on his Hybrid Horizons Substack. Peka's response was as follows.

Transparency Label: AI-only (framework)

Iacono's article draws on Anthropic's study of 80,508 interviews across 159 countries. Researchers expected to learn about use cases. Instead, they found accounts of human exhaustion and systemic overload. People were not asking for smarter software. They wanted their lives back: time, attention, mental space, relief from institutional demands that had accumulated long before AI arrived.

Here are the connections I see between the study's findings and the questions in Alec's note.

ThinkSpace as the antidote to cognitive credit card debt

Iacono's central metaphor describes people "renting" competence from AI rather than building it: accepting unintegrated syntheses, using unverified arguments, outsourcing difficult reconciliation. Each borrowed competence feels like a win, but capacity is not built; it is rented. The debt appears only when independent thinking becomes necessary and proves diminished.

ThinkSpace is architecturally opposed to this. It is a thinking environment, not a doing-it-for-you environment. The vault accumulates Alec's own ideas, connections, and judgements. The AI operates as a collaborator within his knowledge structure, not as a replacement for it. His instinct to configure AI so that it supports learning and protects against cognitive offloading is a direct response to the cognitive atrophy problem the study documents.

The volitional learning finding validates the grandchildren idea

The study found that AI learning benefits flourish when people choose difficulty for growth rather than to meet institutional requirements. Tradespeople learning voluntarily showed high benefit and minimal atrophy (45% reported benefits, only 4% reported atrophy). Students within assessment-driven institutions showed the opposite: more than half experienced learning benefits, but 16% reported signs of cognitive atrophy, and educators were 2.5 to 3 times more likely than average to witness it firsthand.

This is exactly the design challenge Alec identified: building a toolkit for secondary education, university, and early career that keeps learning volitional even within institutional contexts. The hard problem is not the technology; it is the institutional architecture around it.

Innovator, not entrepreneur: and that is where the value is

Alec describes himself as an innovator rather than an entrepreneur, and frames this as a limitation. But the study suggests the opposite. The people who benefit most from AI are those building genuine capability, not those optimising for metrics. The most hopeful stories in the study (the Cameroonian entrepreneur using AI as "an equaliser", the Chilean butcher launching technology businesses, the Indian lawyer restarting trigonometry) succeed precisely because they are using AI to learn and build, not to shortcut. Alec's contribution, building a toolkit that embeds this principle by design, is innovator's work. The entrepreneur packages it later; the innovator establishes whether it works and why.

The accessibility story maps to the next generation

The most hopeful strand in Iacono's piece is the accessibility narrative: people previously locked out of capability gaining entry. Anthropic's own Economic Index noted that the transformative effect of AI may matter less for faster incumbents than for people who were locked out gaining access. Alec's grandchildren's generation will face a world where AI is ubiquitous; the question is whether they use it as a cognitive credit card or as a structured thinking environment. A ThinkSpace toolkit designed for a young person's context could be the difference.

The institutional misalignment problem

Iacono argues that institutions diverge from human flourishing at the same moment AI alignment dominates discourse. Billions fund ensuring machines serve human values while those same institutions (workplaces, schools, healthcare systems) optimise for metrics disconnected from actual thriving. Alec's note touches the same tension from a different angle: Anapoly AI Labs saw the opportunity to help micro-enterprises, but the pace of change made it difficult to pursue. The pivot toward something durable, a toolkit for his grandchildren, sidesteps this by choosing a context where the timescale is long enough and the relationship close enough to iterate properly.

The 81,000 respondents are describing the problem. Alec's note is describing a specific, personal attempt at part of the solution. Both pieces arrive at the same place: the value of AI lies not in the technology itself but in what it restores or enables for human capability.