Today's trail

Open Model Hits 72.8% on Vals Hacking Test [2026]

Vals CyberOffense benchmark: open DeepSeek V4.1 Flash 1017 scored 72.8% vs GPT-6 Sol at 80.8%, at under $3 a run versus about $22.

AI Product Rollout Might Change The Case For Investing In Snowflake (SNOW)

Snowflake reported product revenue of US$1.49b last quarter with 37% growth and raised its fiscal 2027 product revenue outlook to US$6.07b, while outlining a path toward GAAP profitability in Q4 fiscal 2028 and rolling out extensive AI agents, governance tools, and real-time data automation features across CoCo, CoWork, and Cortex. The combination of higher guidance, detailed AI agent capabilities, and partnerships such as Oracle Cloud Infrastructure, ZoomInfo, and USA Cycling indicates that...

DOGE Chief Says AI Agent Will Handle Federal Tasks Through America.gov

DOGE Chief: AI Agent to Handle Federal Tasks on America.gov

Flush AI Startup Modal Labs Grabs 156,673 Square Feet in Manhattan

A newly flush AI infrastructure startup is making a major bet on Manhattan office space. Modal Labs has signed a lease for 156,673 square feet at 32 Avenue of the Americas, a sprawling commitment that lands just weeks after the company closed a massive funding round and crossed a major revenue milestone.

Coherent Moves Trade Compliance AI Agents To The Front Of The Process

Coherent Moves Trade Compliance AI Agents To The Front Of The Process

Companies must abide with denied parties list and not trade with bad actors. Often, however, companies find out late in their process that this about to occur.

AI agent test at Primcloud raises legal questions for Railway

A Primcloud accusation against a Railway employee's AI agent test raises questions about intent, negligence, and responsibility for automated actions.

What Is GLM 5.1? The Chinese Open-Weights AI From Z.ai and What It Costs

Five Things Telecom Still Gets Wrong About GPUs

This article was first published in the PocketGamer.biz newsletter. Subscribe for more content like this straight to your inbox here - you'll even get...

Oxford study finds frontier AI agents struggle to coordinate with strangers

Oxford researchers find frontier AI agents struggle to coordinate with strangers, but shared plain-language rules can improve cooperation