The impact of AI on science is an augury of everything, from a new paper:
- Publish 3.02x more papers
- Receive 4.84x more citations
- Shrinks the volume of topics studied by 4.63%
- Decreases scientists’ engagement with one another by 22%
https://www.nature.com/articles/s41586-025-09922-y
Data center related updates pointing in the same direction and which will be misunderstood:
• Powered land company backs out of Wisoncsin data center https://www.datacenterdynamics.com/en/news/cloverleaf-backs-out-of-data-center-project-in-greenleaf-wisconsin/
• CoreWeave shareholder sues alleging company mislead over ability to deliver compute https://www.datacenterdynamics.com/en/news/coreweave-shareholder-files-lawsuit-alleging-company-mislead-investors-over-ability-to-deliver-compute/
Climate change has already lowered U.S. incomes by ~12% vs what might otherwise have been the case, according to a new econometric study. With AI's sharply growing role, we should look forward to more such losses in future.
https://www.pnas.org/doi/10.1073/pnas.2504376122?__cf_chl_tk=dvW8_nE..ADlGZOTu56fxiqpqHrjpKgNdXYO6_BS6fY-1768414370-1.0.1.1-z46JdO152VzUa5XuKDtLbtq7wjlF.Xu1S9_uRoPtBl8
Heroic stuff from a columnist in the NYT: "My bet is that over the next 18 months, OpenAI runs out of money." https://www.nytimes.com/2026/01/13/opinion/openai-ai-bubble-financing.html?smid=nytcore-ios-share
Earnings and market cap concentration in the S&P 500 are both at all-time highs, driven largely by tech equities, and, more specifically, by AI. We long ago eclipsed the dot-com peak.
Agentic Ai is as prone to conformity and group-think as are groups of humans, but does so even faster, but without human accountability or self-awareness. The intelligence of the AI has no bearing, nor does the visibility of the work. It is a deep, structural flaw and source of risk. https://arxiv.org/pdf/2601.05384
Perhaps less well known than it should be, and it is somewhat masked by the current AI fervor, but tech is a largely mature industry. Microsoft turned 50 last year. The sector's tepid recent job growth reflects that late middle-aged status.
One way to think about the AI paroxysm is aging hipsters trying to recapture their (economic) youth. It rarely works.
AI as productivity tool cited extensively in new Senate report on UHC up-coding in Medicare Advantage. Except the productivity flowed from, for example, using AI to infer likely diagnoses from complex non-clinical signals, pushing that into clinician workflows, and overbilling with abandon. AI, it's the future of medicine. https://www.grassley.senate.gov/imo/media/doc/uhg_report_-_final.pdf
Researchers turn LLMs from AI therapists into patients, and discover that they are, in short, deeply troubled. They shows strong signs of severe depression, defiance, anxiety, and trauma-related shame. It is just what you'd want in our future overlords. https://arxiv.org/pdf/2512.04124
As I say too often, given long-run climate series, this year will almost certainly be one of the three or four coolest of the rest of your life, _and_ that will be true every year for the rest of your life. Be very long central air, and short providers of flood insurance. Meanwhile, it will take a lot of AI to "solve" climate change, given AI's energy profligacy's growing role in climate change.
