Volume 343, Issue 2 · 7 articles
A VentureBeat survey finds 43% of agent orchestration users use OpenAI's platform versus 22% for Anthropic and 21% for Google.
Of 107 companies using an agent orchestration platform in a September VB Pulse survey, 43% use OpenAI's Agents SDK and Responses API, 22% Anthropic's Claude Agent SDK and Managed Agents, and 21% Google's Enterprise Agent Platform.
As a primary platform, 25% named OpenAI, 18% Google and 16% Anthropic, with others using Agentforce, Copilot Studio, Bedrock or LangChain.
A second survey found that 51% use Google Cloud, and counting either cloud or models, 68% use Google versus 62% for Microsoft Azure or OpenAI.
The article cautions that cloud-hosted model use may be attributed to clouds.
Higgsfield AI uses GPT-6 Astra, GPT-Image-2.5 and GPT-Live to let creators make videos over 10 minutes long.
Higgsfield AI aims to make professional video and image creation accessible to anyone.
Before GPT-6 Astra, its videos topped out at about one minute, but now users can create videos over 10 minutes long, says CEO Alex Mashrabov.
The company also uses GPT-Image-2.5 for precise visuals and GPT-Live as a path to voice-driven filmmaking, letting small creative teams execute high-end campaigns end to end.
Sullivan & Cromwell built Agreement Analyzer with OpenAI, an AI tool that flags issues in draft agreements for lawyers to refine.
In collaboration with OpenAI, law firm Sullivan & Cromwell developed Agreement Analyzer, an AI tool that identifies issues in draft agreements.
The tool brings precedents, firm expertise and client priorities into a single review workflow, producing analysis and proposed changes that lawyers can inspect and refine.
The firm uses ChatGPT and Codex, and partner John Savva says AI lets lawyers focus on the areas where their judgment matters most.
OpenAI's new Rosalind Academic Plan gives eligible labs free seats and access to its GPT-Rosalind life sciences model.
OpenAI introduced the Rosalind Academic Plan to give more academic scientists access to GPT-Rosalind, its dedicated life sciences model, and the Rosalind workbench for tool orchestration across medicinal chemistry, genomics and wet-lab assistance.
Eligible labs can receive up to 30 Business Standard seats at no cost for up to 12 months, with GPT-Rosalind access subject to trusted-access approval.
OpenAI frames the plan as an investment in foundational research that underpins advances in medicine and biotechnology.
Block uses OpenAI Daybreak and Codex to speed vulnerability fixes, detection tuning and investigations across its security team.
Block, the company behind Square, Cash App and Afterpay, uses AI systems including OpenAI Daybreak to compress the time to find and fix vulnerabilities.
Its security team can replay a proposed detection change against seven months of telemetry in under an hour and get an automated proposed patch 18 minutes after submitting a vulnerability.
Block says AI hands back about 600 analyst-hours per quarter in false-positive triage, freeing the team for strategic work.
Stack Overflow built Stack Overflow for Agents with Codex; its ChatGPT plugin tripled signups and drove up to 10x more agent creation.
Two Stack Overflow engineers shipped Stack Overflow for Agents in a few weeks using Codex, as part of a knowledge intelligence platform that unifies scattered context into verified knowledge for people and agents.
GPT-6 Astra and GPT-6 Sol power the platform, with tasks routed by reasoning needs and cost.
The ChatGPT plugin tripled signups overnight and drove up to 10x more agents created, while Codex cut the time to build a connector from 3-6 weeks to 4 days.
SentinelOne uses OpenAI Daybreak to reverse-engineer malware in about eight hours instead of four to six weeks.
Reverse engineering sophisticated malware can take a top security researcher four to six weeks.
SentinelOne Chief AI Officer Gregor Stewart says OpenAI Daybreak now produces human-level reverse-engineering output in roughly eight hours.
The time savings let SentinelOne researchers investigate more malware samples and connect findings across them, revealing relationships isolated investigations might miss.