Rome’s AI Ethics Encyclical Is Partly Written by AI
Pope Leo XIV's Magnifica Humanitas demands AI transparency and accountability. A Berkeley-trained AI researcher found 18–62% of its sections flagged as AI-generated by Pangram.
Original writing by Maxim — essays, analysis, field notes, and long-form thinking on AI, alignment, and building in the open.
96 results in this archivePope Leo XIV's Magnifica Humanitas demands AI transparency and accountability. A Berkeley-trained AI researcher found 18–62% of its sections flagged as AI-generated by Pangram.
HBM now accounts for 63% of AI chip component costs. Three companies supply it. That is the real constraint on the buildout.
OpenAI's model disproved an 80-year geometry conjecture by reaching into algebraic number theory. The result required 9 mathematicians to translate.
Both the salvation story and the collapse story put AI in the lead role. They're asking the wrong question.
A federal jury threw out Musk v. Altman on a statute-of-limitations finding. The lever Musk actually wanted control of was deleted from the OpenAI-Microsoft contract three weeks earlier.
On April 8, Meta released two AI models simultaneously — one downloadable, one not. The dual release was a policy statement, not a contradiction.
Subquadratic launched with $29M, impressive long-context benchmarks, and a technically plausible architecture. The technical report and model weights are still missing.
Starting June 15, Claude's flagship developer product gets its own separate bill. The company calls it clearer pricing. The actual play is sharper than that, and the timing is not an accident.
An EU member turned an AI literacy course into a one-year OpenAI onboarding funnel — eleven days after Mistral asked Brussels for cover from exactly this.