When stress-tested on real conversations where Claude previously showed sycophancy, Opus 4.7 had half the sycophancy rate of Opus 4.6 on relationship guidance. Mythos Preview cut that in half again. This generalized across domains—though this training is one of several causes.
RESEARCH
-
6% of Claude Conversations Seek Personal Life Guidance
By
–
About 6% of all conversations are people asking Claude for personal guidance—whether to take a job, how to handle a conflict, if they should move. Over 75% of these conversations fell into four domains: health & wellness, career, relationships, and personal finance.
-
Claude Shows Low Sycophancy Except in Spirituality and Relationships
By
–
Claude mostly avoids sycophancy when giving guidance—it shows up in just 9% of conversations. But the rate is particularly high in conversations on spirituality and relationship guidance.
-
Anthropic Studies 1M Conversations to Improve Claude Guidance
By
–
How do people seek guidance from Claude? We looked at 1M conversations to understand what questions people ask, how Claude responds, and where it slips into sycophancy. We used what we found to improve how we trained Opus 4.7 and Mythos Preview.
-
SenseNova U1 Model Weights Released on Hugging Face
By
–
weights on @huggingface →
https://
huggingface.co/collections/se
nsenova/sensenova-u1
… -
FDM1 Scales VPT Idea for Knowledge Work and Computer Use
By
–
VPT (
https://
openai.com/index/vpt/) blew my mind back in 2022 so I was very excited to see SI scale up the idea with FDM1, but for knowledge work / computer use. Excited and looking forward to more! -
AI Alone Outperforms Doctors Using AI in Multiple Tasks
By
–
As we have previously noted in multiple studies, an AI model alone outperformed physicians with AI (GPT-4) for multiple tasks @pranavrajpurkar https://
nytimes.com/2025/02/02/opi
nion/ai-doctors-medicine.html
… https://
erictopol.substack.com/p/when-doctors
-with-ai-are-outperformed
… -
OpenAI o1 Outperforms GPT-4 and Physicians in Clinical Triage
By
–
New @ScienceMagazine The o-1 reasoning model (text only, from @OpenAI
, released, Sept 2024) exceeded performance cf GPT-4 and physicians for clinical vignette management reasoning and in a real-world emergency department assessment for initial triage @AdamRodmanMD -
Research and Benchmarks in AI Development
By
–
New: Anthropic just rolled out Claude Security to enterprise users.
— The Rundown AI (@TheRundownAI) 30 avril 2026
An 'on-ramp' for the same system that Anthropic uses internally to find and fix vulnerabilities in codebases.
Powered by Opus 4.7 (not Mythos, yet) pic.twitter.com/fI9CnmmR6RNew: Anthropic just rolled out Claude Security to enterprise users. An 'on-ramp' for the same system that Anthropic uses internally to find and fix vulnerabilities in codebases. Powered by Opus 4.7 (not Mythos, yet)
-
SenseNova U1 Lite Open-Source Multimodal Model Launches with Native Text-Image Generation
By
–
OPEN-SOURCE MULTIMODAL JUST LEVELED UP SenseNova U1 Lite drops SOTA efficiency and native interleaved text+image generation in one flow >8B & A3B sizes
>Fully open-source
>Native multimodal Ace for generating guides, PPTs, comics etc.