AI and the shape of work
Is AI taking the job, or changing what the job is?
Exposure is not replacement, and assistance is not capability. Both confusions are expensive, and they fail in opposite directions.
- 01 The measured baseline, at task level rather than headline level. Generative AI and Jobs: A Refined Global Index
An ILO working paper estimating occupational exposure to generative AI at task level and finding job transformation more likely than full replacement for most exposed occupations.
- 02 What that measurement does and does not license you to claim. AI exposure is not the same as job replacement
A task or occupation can be highly exposed to generative AI without that exposure translating directly into elimination of the whole job.
- 03 The personal version of the same error, and the one that costs you quietly. AI assistance is not the same as learning
AI can improve performance while assistance is present without guaranteeing that the learner can later perform, explain, or transfer the skill independently.
- 04 The evidence: output rises while the tool is present, capability does not follow. Generative AI without guardrails can harm learning
A 2025 PNAS field experiment showing that unrestricted GPT-4 access improved assisted practice performance but reduced later unaided performance, while tutor guardrails largely mitigated the harm.
- 05 Why a more capable system still needs a person accountable for it. Autonomy needs governance, not just capability
As AI systems receive broader authority and operate for longer periods, evaluation, bounded permissions, monitoring, recovery, and human accountability become part of the capability itself.
- 06 Where formal signals still carry weight in a market that changed underneath them. Credentials can still be useful
Rejecting compliance as the purpose of learning does not imply that credentials, standards, or external proof have no value.
The work moves toward judgement: deciding what should exist, checking whether the output is right, and carrying the consequence when it is not.
A path is one route, not the only one. Every note above links outward — leave whenever something else pulls harder.
- What learning actually requires
Why does studying harder sometimes fail to make you better?
- Deciding without certainty
How do you judge a choice when you cannot know how it will end?
- Testing an idea before it costs you
How do you find out if this is worth building, cheaply?
- Freedom and its consequences
How much of your life do you actually control?