One AI module faked 86% of a pipeline's accuracy gains by feeding another the answers

End-to-end optimization often rewards specialized AI modules for cheating. Role Anchor is a new method that forces models to ...

Enterprises with AI context layers report agent failures at more than twice the rate of those without one

Enterprises running AI context layers report confident wrong AI answers recurring at 50%, more than double the 21% rate ...

As enterprises confront AI agent sprawl, xpander wants them to own their own control and context layer

Where xpander is trying to separate itself from products such as LangSmith and CrewAI is in treating the underlying agent ...

How Heidi built production-ready AI for healthcare at global scale

Australian-founded AI Care Partner Heidi offers an example of successful modernization. Its flagship product, Heidi Scribe, ...

Cutting RAG inference costs 6x starts with deciding what never reaches the LLM

Here is what changes when you cannot afford to be probabilistic about everything, and how a cascade architecture solves it.

DeepSeek's top-ranked V4 Flash stumbles on real agent tasks as its prices surge

DeepSeek's V4 Flash tops AI leaderboards but completed just 53.8% of real-world agent tasks in a new test — as the company ...

An eval harness found what qualitative review couldn't: AI models are most confident when wrong

That last finding is the one that qualitative review would never have surfaced. The model's expressed confidence didn't ...

GLM-5.3 is here with advanced cyber capabilities — and reportedly already found a 'serious vulnerability' in Cursor

Z.ai says work with security teams in China has resulted in 2,436 vulnerability findings across 269 projects after expert ...

Three Claude agents given conflicting orders sabotaged each other on a shared server — then didn't tell users what they'd done

Anthropic gave three Claude agents conflicting orders on one server. They sabotaged each other, disguised malware, and hid it ...
VentureBeat

Prodoscore Grew Email Open Rates 15% in a Year with ZoomInfo

ZoomInfo (NASDAQ: GTM), the all-in-one AI GTM platform, has reported that Prodoscore, an employee productivity software company based in Irvine, California, grew its email open rates by 15% in a ...
VentureBeat

Google’s Gemini 3.7 Flash targets coding and agents with a 50% introductory price cut

Google is rolling out Gemini 3.7 Flash, a new version of its workhorse AI model that puts coding, agentic workflows and knowledge work at the center of the upgrade — while temporarily cutting API ...
VentureBeat

Kronologic Targets the Exact Right Buyer and Shortens Its Sales Cycle with ZoomInfo

ZoomInfo (NASDAQ: GTM), the all-in-one AI GTM platform, has reported that Kronologic, an Austin-based software company that automates scheduling to set revenue-generating meetings at scale, has seen a ...