Models

New AI models and products, what they can really do, and how they compare.

A lone Android phone displays an unbranded personal app, with no public store listing visible.

Explainer

Google Play’s vibe-coded apps: what to know about personal use and public release

Google says the rollout starts with adults in the U.S. and selected markets. Friend-and-family sharing is planned; public distribution still requires a professional account and reviews.

Models Desk10 hr ago4 min read

Thumbnail of the YouTube video “I Tested the Viral Reverse Engineer Anything Skill. What You Need to Know.”

Analysis

Nate Herk | AI Automation says Reverse Engineer Anything helps nonprogrammers build more professional-feeling projects

Nate Herk | AI Automation says the idea is not new: Reverse Engineer Anything’s value is a practical framework for studying examples and rebuilding specific features.

11 hr ago

Google rolls Gemini 4 Argon into its app for Pro and Ultra subscribers
Artificial Analysis and Vals AI provide benchmark results, but their scores do not establish that Argon can reliably solve an individual user’s hardest real-world task.
13 hr ago
Explainer
Meta’s Muse Wonder frontier claim: what to know
Wonder Full scored 56 on Artificial Analysis’ index, below Claude Opus 5.5’s 58; Meta’s task-specific claims and safety assessments have not been independently established.
Yesterday

Analysis
Bijan Bowen sees no reason to choose Step 5 Preview over DeepSeek Flash
Bowen was fairly disappointed by recurring spatial problems and saw no reason to choose Step 5 Preview over DeepSeek’s similarly sized Flash model.
Yesterday
Europe's AI independence takes more than open models, despite Tsai's claim
EU plans also target computing, investment, skills and industrial adoption. Its proposed gigafactories have yet to be selected, and expected investment is not money already spent.
Oct 10
What is OPM’s FWD Chat beta meant to do with federal workforce data?
The Oct. 7 beta lets users inspect data filters and time periods, but OPM says important figures should be checked elsewhere; an America.gov connection remains a plan.
Oct 9
Mistral Large 4 trails four Chinese open-weight models on a composite index
The index favors four Chinese models, while separate legal and cybersecurity evaluations give Mistral’s preview stronger results on particular tasks.
Oct 9