2026-05-24
One simple message for the day
Skill with AI is something you build by using it, not something you're born with. The people getting the most out of these tools are mostly the people who have spent the most hours on them — which means the gap is learnable, not fixed.
Eight bullets
-
01
The useful skill isn't "knowing AI." It's knowing your own work well enough to tell when the machine got it wrong.
-
02
Anthropic looked at two million real chats: longer-term users had roughly a 4-point higher success rate, even on the same task. Practice shows.
-
03
A study of 10 million UK job ads found AI skills carried a 23% pay premium — higher than a Master's (about 13%). One country, one dataset, but a striking gap.
-
04
AI roles in that data were about twice as likely to offer parental leave and three times as likely to offer remote work. The competition is over conditions, not just pay.
-
05
On résumés, listing AI skills helped older applicants and people without degrees the most. A small equaliser, not a divider.
-
06
Coding is quietly leaving the chat window and moving into the pipes — fewer people "asking" an AI to code, more code running automatically in the background.
-
07
Half of all jobs have now had at least a quarter of their tasks touched by AI in real use. Touched, not taken.
-
08
The honest version of the AI story this week: the tool is ordinary now. The skill is in deciding what's worth doing at all.
Worth reading
Per house rule: any figure about jobs and AI is one study's finding, not a settled fact.
These 3 charts show how AI is affecting wages, job quality and hiring decisions
Clearest primary write-up of the AI wage premium and the hiring-experiment evidence.
Economic Index report: Learning curves
Real-usage data, not surveys. The "learning by doing" finding is the day's spine.
AI transformation is reshaping work. HR leaders must help redesign it
The redesign-of-work angle, written for people who actually run teams.
Agents, human agency, and the opportunity for every organization
Vendor source (read with that in mind), but useful on where human judgement stays.
Maps which tasks workers actually want to keep. The "human agency" data is good.
Bridging Skill Gaps for the Future: New Jobs Creation in the AI Age
Counterweight to the doom framing — looks at jobs created, not only lost.
AI and jobs: a review of theory, estimates, and evidence
A sober survey of 400-plus studies. Good for checking any scary single number.
Complement or substitute? How AI increases the demand for human skills
Evidence that demand is shifting toward creativity and problem-solving. Preprint — not peer-reviewed yet.
How Artists Actually Use AI in 2026 and What That Means for Music
The creative-practice view — artists using AI for parts of a song, not the whole.
Millions of people are pretending to be AI chatbots — for fun
The fringe/funny one. Humans doing an impression of the machine that's imitating them.
Bosses are becoming obsessed with ChatGPT. It's creating more work for employees
The necessary cold water. Adoption without a plan can add work, not remove it.
— Peter Drucker · Fits the day: AI makes doing things faster; it doesn't decide whether they're worth doing.
Give the machine the task you know best — then study only what it left out.