@AYi_AInotes: Mind blown—22-year-old founder open-sources what Anthropic kept locked away: the Claude Mythos black box
Summary
A 22-year-old founder claims to have cracked open the architecture black box of Anthropic’s Claude Mythos model via an open-source project, speculating it uses a recurrent-depth Transformer design instead of simply scaling parameters.
View Cached Full Text
Cached at: 04/21/26, 07:55 AM
After reading this open-source project, I was completely stunned. The black box of Anthropic’s closely guarded Claude Mythos model has been cracked wide open—and fully open-sourced—by a 22-year-old founder.
This 22-year-old AI entrepreneur tore the lid off Claude Mythos’s black box. He believes Anthropic’s latest god-tier model isn’t the result of simply stacking more parameters, but instead relies on an architecture called Recurrent Depth Transformer.
Similar Articles
@IndieDevHailey: 22-Year-Old Reverses Anthropic's Black Box in Two Days! Claude Mythos Architecture Fully Open-Sourced Anthropic Unveils Most Dangerous AI—Claude Mythos, Capable of Autonomously Uncovering 20-Year-Old Zero-Day Vulnerabilities in OS and Browsers, Only Available in Project Glasswi…
22-year-old developer Kye Gomez reversed Anthropic's Claude Mythos black box architecture in just two days and open-sourced the OpenMythos project, using Recurrent-Depth Transformer and other techniques, achieving performance equivalent to a 1.3B model with 770M parameters.
@seclink: Anthropic released Claude Fable 5 on June 9. As a “Mythos”-class model, it leads benchmarks in coding, research, and visual processing, especially good at large-scale projects like code migration. The model is now available on http://Claude.ai…
Anthropic released Claude Fable 5 on June 9, a “Mythos”-class model that leads benchmarks in coding, research, and visual processing, especially good at large-scale projects like code migration. It is now available to Pro, Max, Team, and Enterprise users.
@FakeMaidenMaker: If you really want to understand how the underlying of AI Agents like Claude Code is built, this open-source project writes one from scratch for you to see. GitHub has garnered 66.5K Stars, and also made it to the Trendshift hot list. The intelligence of an Agent comes from the model itself; what you can do is not "building intelligence"...
An open-source project teaches you to build a simplified version of Claude Code from scratch, thoroughly explaining the harness engineering of AI Agents. It has received 66.5K Stars.
@ruanyf: Anthropic founder's recent remarks: AI open source is a false proposition. He says that even if an AI model is made public, you can't see its internal workings, so the industry doesn't call it 'open source' but 'open weights'. It's not the traditional open source model either; only the weight files are released, and you can't see the model's inner workings, and you can't submit modifications or participate in its…
Anthropic founder Dario Amodei believes that AI open source is a false proposition because only the weights are released, not the source code, so users cannot participate in modifications. Blogger Ruan Yifeng criticizes this view as biased, pointing out that open source models still have advantages in privacy and controllability, while also accusing Anthropic of discriminatory account bans against Chinese users.
@AlchainHust: https://x.com/AlchainHust/status/2064676532212097418
This article provides a detailed review of Anthropic's newly released Claude Fable 5 model, and demonstrates the author's process of using it to develop a Mac App '翻箱' in one day. The model has significant improvements in code generation and stability.