Anthropic acquired the creators of Bun, Claude failed to recreate Space Jam's 1996 website, and Google reversed a product kill. Those are three of the five stories Jerod Santo covers in this Changelog newsletter edition, and each one lands differently than the headline suggests.
The anchor argument is about AI needing hard rules, not intuition-based guardrails. The framing: confident systems that are wrong are more dangerous than uncertain ones that say so. The Anthropic-Bun acquisition thread pulls that further, raising the question of what it means when AI companies start absorbing the infrastructure layer.
The Space Jam failure is not a throwaway bit. It is a concrete, reproducible test of whether current models can faithfully reconstruct a known artifact from the past, and the answer is no. That detail, combined with the rules-vs-vibes argument, is why this one is worth reading past the summary.
[READ ORIGINAL →]