Former Anthropic Researcher Says Algorithmic Efficiency Is Accelerating, Left Over Chart Dispute
Understand the story
Former OpenAI and Anthropic researcher Jacob Coxon said on November 14, 2026 that a chart shown during Anthropic's IPO roadshow, depicting the growth of its pretraining compute multiplier over time, was a major reason he left after about four months on the job. Coxon also claimed that Anthropic could retrain its Mythos model with only a small fraction of the compute, citing this as evidence of accelerating gains in algorithmic efficiency. The report so far only covers Coxon's statements and his reasons for leaving; Anthropic's response or any follow-up developments were not mentioned.
AI overview of reporting · Updated 18 h前
Reporting timeline
Explore different aspects of the story through its reporting.
- The DecoderAnthropic can already retrain Mythos with "a fraction of the compute originally needed," says former researcher Coxon
Jacob Coxon, a former OpenAI and Anthropic researcher, said a chart shown during Anthropic's IPO roadshow depicting the pretraining compute multiplier over time was a major reason he left about four months after joining.
Story heat trend
Current heat6·Comparable peak10（11/14 20:00）·Comparable change over 24 hours –
The trend compares the same continuously observed sources, which may cover fewer sources than the current heat score. Hover or select the chart to inspect each hour; use the left and right arrow keys to navigate.