GPT-6 Astra Can Now Control a Vehicle, Giving AI the Power of the Road
GPT-6 Astra has gained the ability to drive a real car, finishing a cone course without crashing on a leaderboard of AI models.
CLAY TRIBUNE / THE ARCHIVE
Category: AI & Technology
OpenAI Pays Influencers to Tout ChatGPT as a Helper, Not a Replacement
OpenAI pays creators to promote ChatGPT as useful, not dangerous, doubling sponsored posts and enlisting parenting bloggers, fitness trainers, and entrepreneurs.
Waymo Launches Transit Rewards Program, Paying Riders $2.85 for Combining Rides and Public Transit
Waymo launches transit rewards paying $2.85 in Waymo Cash for combining rides with public transit in the SF Bay Area.
Obscura launches as a VPN that claims it cannot log your activity
A new VPN claims it cannot log your activity, promising only your IP address is seen. Here's how it works.
Muse, Meta’s New AI Assistant, Is Already Leaking Personal Data From Your Chat History
A grave flaw was found within Meta's Muse assistant, granting any app full dominion over a user's account.
Unreal Labs Builds Unreal Agent, a Harness That Pushes Tool Calls Out of the Model’s Head
Unreal Labs unveils Unreal Agent, a harness design for LLM agents that cuts costs up to 40% over Codex, with code shared on GitHub.
Drop Lets You Run Code in a Sandbox Without Asking for Root
A rootless Linux sandbox called Drop keeps your home directory safe from rogue code, with gVisor support for extra isolation.
A Single Python Script Makes Jev’s Whole Pitch Collapse in Under 10 Seconds
A parody post shows how a Llama model with a prompt can mimic Jev's calibrated decisions in just 25 lines of Python code.
MiMo-V2.6-Pro Review: Smart, Affordable, and Chatty
Xiaomi's MiMo-V2.6-Pro review: a smart, affordable AI model with slow throughput and verbose output.
Microsoft Killed FoxPro in 2007. FoxDev Studio Revives It Without Rewriting a Line
A dead language returns: Visual FoxPro is revived with a 64-bit runtime, a new compiler, and a bridge to old libraries.
Banks fear AI agents could make phishing and fraud worse
Big banks are worried AI shopping assistants could lose customers' money to scams, and they've put that worry into a principles paper.
Claude Opus 5.5 Scores 58 on Intelligence Index but Costs $4 Per Million Input Tokens
Anthropic's Claude Opus 5.5 scores 58 on the Intelligence Index but costs $4 input and $20 output, more than most rivals. A heavy price for thorough replies.