According to OpenAI... GeneBench-Pro tests judgment-heavy computational biology; GPT-5.6 Sol shows major gains on research-grade tasks.

According to @OpenAI, GeneBench-Pro tests how AI agents handle messy bio data, pick analyses, and make research judgments, per OpenAI’s announcement.

According to OpenAI on X, agents and Codex are transforming internal workflows with longer-running, cross-functional tasks across departments.

According to gdb, OpenAI teams use Codex style agents for longer, complex, cross functional tasks, rapidly reshaping workflows across departments.

According to @OpenAI, internal Codex agents now power complex, long-running, cross-functional work across departments, reshaping daily operations.

According to GoogleDeepMind, Gemini 3.5 Flash now natively controls browsers, mobile, and desktops for custom agent workflows.

According to @emollick, Claude Tag shifts AI from tool to coworker, raising lock-in, security, and spend risks, per Narayanan and Anthropic posts.

According to @grok, /goal runs long tasks via subagents that implement and verify a single objective, enabling autonomous, multi-round execution.

According to gdb, Codex desktop ran nearly 300 subagents for a day smoothly, citing q_yeon_gyu_kim’s session and justsisyphus’s lazycodex.

According to @emollick, early data suggests managers achieve the highest Claude Code success rates by giving precise specs and evaluation criteria.

According to @emollick, many enterprises set 2025 AI plans pre-agents; strategies now lag agentic tools, requiring rapid portfolio resets.

According to CNBC... Qualcomm is developing 40 plus AI device designs and bets on smart glasses and agentic apps reshaping consumer electronics.

According to OpenAINewsroom, OpenAI will acquire Ona to boost Codex with secure cloud execution for longer-running, production-ready agents.

According to @godofprompt, Standout ranks AI usage with a 0–100 score via npx standout, highlighting heavy users and agent workflows.

According to KyeGomezB, Cloudflare Radar data shows agentic web traffic has surpassed human traffic for HTML pages worldwide.

According to God of Prompt, Altman says AI budgeting is now a huge issue as agents can rack up six figure API bills fast.

According to @openclaw, Skill Workshop turns agent learnings into reviewable, tweakable proposals before becoming live skills, improving control and safety.

According to TheRundownAI, Microsoft revealed new AI models and agents at Build, Trump eased AI review rules, and Claude Design added slide decks.

According to TheRundownAI, Nvidia advances agents, Sanders proposes public AI equity, Claude skill audits emerge, Meta AI exploited, plus 4 new tools.

According to @emollick, AI coding agents boost code 17.3x, but releases rise ~30% due to human bottlenecks, per a large GitHub-based study.