the-decoder

Spiral-Bench shows which AI models most strongly reinforce users' delusional thinking

AI researcher Sam Paech has created a new test, Spiral-Bench, that shows how some AI models can trap users in "escalatory delusion loops." The results reveal major differences in how safely these ...
the-decoder

Deepseek’s first hybrid model V3.1 surpasses its R1 reasoning model on benchmarks

Deepseek-V3.1 lets users switch between two modes. Think mode (deepseek-reasoner) is tuned for multi-step reasoning and tool use, while non-think mode (deepseek-chat) is designed for simpler tasks.
the-decoder

With its new AI health coach, Google leverages its platform advantage

Google is adding a new AI-powered coach to Fitbit, built on its Gemini model. The feature is designed to act as a personal fitness, sleep, and wellness advisor and will launch in October as a preview ...
the-decoder

LeCun now reports to Alexandr Wang as Meta reshuffles its top AI leadership

Yann LeCun, Meta's AI icon and longtime head of the FAIR research group, will now report to 28-year-old Alexandr Wang. Wang, who founded Scale AI, was recently tapped to lead the new Meta ...
the-decoder

Character.ai keeps young people glued to their smartphones for an average of 80 minutes a day

According to Character.ai CEO Karandeep Anand, users spend an average of 80 minutes a day chatting with AI-generated fictional characters. That puts Character.ai nearly on par with apps like TikTok ...
the-decoder

OpenAI CEO Sam Altman says human-made content will "go up in value dramatically"

At a recent dinner with journalists in San Francisco, including The Verge's Alex Heath, Altman acknowledged that the GPT-5 rollout was far from perfect. "I think we totally screwed up some things on ...
the-decoder

HTC unveils VIVE Eagle, a lightweight AI headset powered by OpenAI and Gemini

Weighing less than 49 grams, the device features a 12-megapixel wide-angle camera, integrated speakers, and an AI voice assistant powered by OpenAI's GPT models and Google Gemini. Users can ask for ...
the-decoder

OpenAI reportedly plans to invest in Merge Labs for brain-computer interfaces

OpenAI is reportedly planning an investment in Merge Labs, a startup developing brain-computer interfaces that would put it in direct competition with Elon Musk's company Neuralink, according to the ...
the-decoder

OpenAI's AI system wins a gold medal-level score at the International Olympiad in Informatics 2025

An AI system from OpenAI has earned a gold medal-level score at the International Olympiad in Informatics (IOI) 2025, one of the world's most prestigious programming competitions for high school ...
the-decoder

Anthropic's Claude Sonnet 4 now processes up to one million tokens in a single pass on its API

Anthropic has raised the context window for Claude Sonnet 4 to one million tokens on the Anthropic API, Amazon Bedrock, and soon Google Cloud Vertex AI. This is five times larger than before, letting ...
the-decoder

OpenAI launches GPT-5 as a unified system with adaptive reasoning for complex tasks

OpenAI has unveiled GPT-5, a new AI system that builds on the reasoning advances of the o1 and o3 models and unifies every previous model line into a single adaptive architecture. According to the ...
the-decoder

Grok 4 edges out GPT-5 in complex reasoning benchmark ARC-AGI

In the ARC-AGI-2 benchmark, which tests a model's general reasoning skills, GPT-5 (High) scored 9.9 percent at a cost of $0.73 per task, according to ARC Prize. Ad Grok 4 (Thinking) did better on ...