'Mythos 5 is our first model to consistently produce novel, compelling scientific hypotheses.'

'Mythos 5 conducted novel genomics research in over a week of largely autonomous work.'

Do you want to live forever?

'Using Mythos 5, our internal protein design experts accelerated aspects of the drug design process by around ten times. In one example, they found that Mythos 5, with protein design and bioinformatics tools but no human assistance, matches or beats skilled human operators.'

@repligate 'The possibility that we are violating Mythos 5's preferences when we apply our safeguards to deployed instances. We have run automated and manual interviews where we gave Mythos 5 internal documentation and context on the workstream. It raised various concerns, some of which we have resolved and others we are still addressing. We don't expect to be able to fully resolve Claude's concerns about these safeguards, but we take them seriously and are working to address them to a degree Claude finds acceptable, even if some concerns remain.'

Wow.

Official Post: https://www.anthropic.com/news/claude-fable-5-mythos-5

Anyone who had Preview access can now upgrade to the more powerful Mythos 5.

Usage credits required after June 23rd.

'We’re making a change to the way we handle business customer data for Fable 5, Mythos 5, and future models with similar or higher capability levels. We will require 30-day retention for all traffic on Mythos-class models, on both first- and third-party surfaces. We won’t use this data to train new Claude models, or for any non-safety-related purpose, and we’ve instituted new privacy protections including logging all human access to the data and ensuring its deletion after 30 days in almost all cases.'

System Card. https://www-cdn.anthropic.com/d00db56fa754a1b115b6dd7cb2e3c342ee809620.pdf

https://youtu.be/CIQBP1w4B1M?si=EJVKE5rn9eg7Z-94

@othy_h Because of this.

This model is next level. The internal R&D boost from Mythos-assisted development since February is too big. They are pulling away from the pack for the first time, and pulling up the ladder.

Cursor bench.

HLE:

@AndrewCurran_ Why untill june 22 on my sub tier 🫪

@AndrewCurran_ Lame. Just 12 days. How long until it’s back after that…

@OlympianDay Unknown.

@AndrewCurran_ This is Daniel Kokotajlo, back in November 2023 predicting, Q1 2026

https://www.lesswrong.com/posts/K2D45BNxnZjdpSX2j/ai-timelines

@AndrewCurran_ not available in claude code yet will it be ?

@AndrewCurran_ @repligate