Intelligence Track

Design

23 June 2026

The Brief

The same force shows up twice today wearing different clothes: the cost of producing a unit — a seat, a line of code — is falling, and the people who control the unit are deciding to keep the surplus rather than pass it on. Airlines are pocketing the gap between sticky fares and cheaper fuel as margin; engineering orgs are pocketing AI's velocity gains right up until a gutted review function turns surplus into a public failure. What connects them is a question of where the released value goes and what gets hollowed out to capture it — discipline in one frame, dangerous thinning in another. The human-capital read sharpens the edge: when execution gets cheap, the scarce thing becomes judgment about what to keep, what to cut, and what not to cut, and the orgs that strip out the wrong layer learn the difference in production. The thing worth sitting with is that every one of today's efficiency gains is also a decision about what to stop doing — and the failures cluster not around the new capability but around the supports quietly removed to afford it.

Lenny's Newsletter · 23 Jun 2026

An executive coach to frontier-lab leaders argues that as AI commoditizes knowledge and effort, teams are restructuring into 'NBA rosters' — flatter orgs, shrinking headcount, player-coach roles, more capital per person — where the scarce, decisive skill becomes emotional clarity: discernment, productive conflict, willingness to fail, and positive self-talk.

Industry lens

As Indian OTA and product orgs adopt agentic tooling, will the first measurable hiring shift show up as reduced execution-role headcount or as new player-coach and judgment-weighted roles — and which competitors move first?

Reading as
Slow down to speed up: so much has changed in 6 months’ time

A survey of six months of engineering change argues that more capable agents since late 2025 have lifted individual productivity while team productivity stays flat, and uses Meta's account-takeover outage — caused by AI-generated, AI-reviewed code shipped amid gutted integrity and security teams — as evidence that deliberately slowing down can be the faster strategy.

The concrete failure mode named here is AI-authored, AI-reviewed code reaching production with no human in the loop after the teams that would catch it were reassigned — meaning the risk isn't agent capability but the removal of the review and quality scaffolding around it.

Pragmatic Engineer·23 Jun 2026