Anthropic is developing a new artificial intelligence model that can infer a user's underlying goals from minimal instructions, going beyond what is asked to understand what is actually meant, The Information reported on Monday.
A person who tested the model told the publication it can identify the broader problem behind an individual software bug and propose more comprehensive fixes, rather than addressing the symptom alone.
The model also generates more creative solutions than existing systems, according to the report.
It comes amid separate reports that Google is working on Gemini 4 Carbon, an upgrade to Gemini 4 Argon which launched last week, with developers describing its models as particularly strong on visual tasks including video analysis and website testing.
OpenAI, meanwhile, is to readying a new model in its Terra family, expected within weeks, offering improved capabilities at a lower price
Read more
