Frontier models clear self-replication's substeps and fail the loop itself
OpenAI's own test scored GPT-4o zero on the end-to-end loop over 100 tries, and the paper that announced a red line crossed had left the model's weights and launch script sitting in the folder.
A chatbot reads your whole prompt at once but writes its reply one word at a time
The reply costs more than the prompt because generating each word means rereading the prompt's cached keys and values from memory.
Claude Opus 5 publishes five different per-token prices
Run one fixed task on that model and its real cost swings 3.6-fold across caching, batch tiers, and the output tokens the headline number leaves out.
Amazon's resume screener taught itself to penalize the word 'women's'
Everything known about the scrapped experiment traces to a single 2018 Reuters investigation and five anonymous sources, and the flaw it describes now lives inside the resume screeners companies use today.
AlphaFold predicts a protein's shape, not what it does
DeepMind's 2021 paper measured how closely predicted protein shapes match the ones labs solve, and beat every rival on a blind test by a margin worth understanding before the Nobel-season shorthand runs past it.