Jonesing for the Next Disruptor (www.datagubbe.se)

The author frames this piece around the 2025 Nobel in Economics awarded to Joel Mokyr, Philippe Aghion and Peter Howitt for explaining how accumulated scientific knowledge and “creative destruction” drive sustained growth — then asks why Western living standards and public infrastructure feel stagnant despite GDP gains. Citing Sweden’s lost industrial competence (Northvolt), creeping rent-seeking, “bullshit jobs,” and stalled consumer-product innovation, the argument is that knowledge alone hasn’t translated into durable, practical advances. That context sets up a broader cultural and policy obsession: find a single “next disruptor” to rescue growth. AI — and especially LLMs — is presented as today’s all‑in bet. The author details technical and economic frictions: LLMs still hallucinate, confuse facts, and produce flaky summaries; agentic/code-generation tools have produced catastrophic failures (e.g., deleted production data) and, in some metrics, reduced developer productivity rather than raising it. Bold vendor claims (Anthropic’s “90% of code” prediction) haven’t materialized, while unit economics look worrying: tokenized LLM calls are often subsidized by VC cash, with major players unprofitable (OpenAI reportedly lost ~$5B in 2024 despite sky‑high valuations). The takeaway: AI hype risks repeating past bubbles unless realistic measures of productivity, robustness, deployment costs, and manufacturable impact replace wishful thinking.