# Former Researcher Says Anthropic Can Retrain Mythos with Minimal Compute

- Category: Flash (快讯)
- Published: 2026-11-14T17:26:26.070Z
- Source date: 2026-11-14
- HTML: https://www.88z.ai/en/news/news-former-researcher-says-anthropic-can-20261114
- Markdown: https://www.88z.ai/en/news/news-former-researcher-says-anthropic-can-20261114.md
- Other language: https://www.88z.ai/news/news-former-researcher-says-anthropic-can-20261114.md

Former Anthropic researcher Jacob Coxon states that the company's algorithmic improvements are accelerating, allowing it to retrain the Mythos model using only a fraction of the original compute. Citing this efficiency leap as his reason for leaving, he calls for a pause in frontier model development to let competitors catch up and commoditize human-level AI.

## At a glance

- Anthropic can retrain Mythos with a fraction of original compute, signaling accelerating efficiency
- Former researcher Coxon calls for a pause in frontier AI development to foster competition and commoditization

## Accelerating Algorithmic Efficiency

Jacob Coxon, a former researcher at Anthropic, highlights a slide from the company's IPO roadshow to demonstrate that its algorithmic improvements are accelerating. He notes that Anthropic can now retrain its Mythos model using only a fraction of the compute originally required. This development underscores significant progress in reducing training costs and enhancing operational efficiency.

## Call for Development Pause

Coxon cites this rapid technological iteration as his primary reason for departing the company. He continues to advocate for a pause in the development of frontier AI models to allow competitors time to catch up. His ultimate goal is to transform human-level AI into a commoditized resource rather than keeping it monopolized by a few tech giants.

## Sources

- [the-decoder.com](https://the-decoder.com/anthropic-can-already-retrain-mythos-with-a-fraction-of-the-compute-originally-needed-says-former-researcher-coxon) (2026-11-14)

This article is an AI-assisted synthesis by 88Z AI based on the public sources below. Product claims and availability are attributed to their publishers. It does not represent independent testing or first-hand reporting.
