Steve Newman writes Second Thoughts (secondthoughts.ai), where he publishes periodic assessments of the state of AI progress, its drivers, and its measurable effects.
Positions and statements
On AI's measured impact ("Anecdotes Everywhere, Evidence Almost Nowhere", July 2026, the third part of a mid-2026 series). Newman argues that AI has so far had "very little clear impact on the world at large," with the data-center capital expenditure boom as the principal exception — citing a St. Louis Fed estimate of 0.97 percentage points of contribution to real GDP growth over the first three quarters of 2025, with his own caveat that the underlying series covers all information-processing equipment, software, R&D, and construction spending rather than AI alone.
His methodological position is that widely cited AI-impact statistics mostly do not carry information: the Census Bureau's 18%-of-firms adoption figure records only that a firm ticked a box, and macroeconomic series are too noisy to attribute movements to AI. He treats the absence of a clear signal as "suggestive evidence of absence" for the period through mid-2026 while arguing that 2026 may be the last year for which that holds, given AI metrics growing at 3× to 10× per year. His stated prescription is investment in faster data collection and organizational capacity to respond, rather than a substantive policy position.
On labor specifically he holds an explicitly unresolved view: certain of enormous change over time, guessing overall employment will decrease, thinking it possible that entry-level positions are already affected, but finding no clear signal in aggregate employment data. See AI and Productivity, AI Labor Disruption.
Newman's earlier installments in the same series covered the pace of AI progress ("Is AI 2027 Coming True?") and its drivers ("The AI Race"); neither is separately tracked here.
Relationships
- supports: Anecdotes Everywhere, Evidence Almost Nowhere (Steve Newman, July 2026) — author
- related: AI and Productivity, AI Labor Disruption, AI Is Really Weird