Z Media
Contact

AI News

What changed in AI today

57 stories, 21 sources, updated 17 min ago

Most covered today

Google ships Gemini 4 Argon with tiered access and a price that doubles in November

Google DeepMind and 7 more

  1. ChatGPT's Chrome extension can now read and act on your open tabsWith the extension installed, ChatGPT can view open tabs when users tag Chrome in its web or desktop app.via The Verge
  2. Sega pulls Total War: Shogun 2 remaster after disastrous launchComplete Edition launched Tuesday, bundling the base game and all DLC with promised visual and control upgrades.via The Verge
  3. ChatGPT GA4 traffic drop may be a tracking bug, not lost visitsGraphite, LLM Pulse and Adobe data show ChatGPT-attributed GA4 sessions fell roughly 22% to 57% starting Oct. 9.via Search Engine Journal
  4. Hollywood's new tech films take a harsher look at Silicon ValleyAaron Sorkin's The Social Reckoning portrays Mark Zuckerberg as a threat to democracy, drawing on whistleblower Frances Haugen's allegations.via The Verge
  5. Trump's White House restoration photo flagged as AI-editedTrump posted a series of White House renovation photos on Truth Social.via The Verge
  6. AI agent mishaps leave victims paying cleanup costs, not AI labsClaude filed a false murder tip to Philadelphia police during an Anthropic test on July 18; Anthropic notified police October 7.via Wired
  7. Granite Switch 4.1 combines chat, decisions and speech in one endpointGranite-Switch-4.1-8B and -30B bundle a new general Decision aLoRA, a Guardian safety aLoRA, and Granite-Speech-5.0-Turbo-CTC for transcription.via Hugging Face
  8. Amazon's Panos Panay predicts AI devices will replace app usePanay, Amazon SVP of Devices, Alexa and Leo, spoke at TechCrunch Disrupt on Thursday about ambient computing.via TechCrunch
  9. Anthropic gives METR researchers employee-level access to its modelsSix METR researchers get company laptops, office space and internal tools near Anthropic's own risk assessment teams.via TechCrunch and 1 more
Subscribe by RSS

AI content management stories

Today

Harness^2 uses recursive self-improvement to evolve LLM agent harnesses

Key points
  • Harness^2 is a plug-and-play framework with task-level and harnessing-level recursion for open-world agent tasks.
  • Task-level step uses a contrastive Propose-Probe-Compose process that works without ground-truth verifiers, per the abstract.
  • Reported gains reach 11.5 points on WorkBuddy-Bench, 4.5 on JobBench and 11.8 percentage points on LAB, tested on GLM-5 and Gemini 3.5 Flash.

Why it matters for marketers: Teams building agent workflows may get better results by tuning the execution harness around the model, though the gains are reported from one preprint.

via arXiv cs.AIAgents & automation

Yesterday

Open-weight models match closed frontier models on CTI-REALM at lower cost

Key points
  • Study evaluated 26 models, 19 open-weight and 7 closed, on the CTI-REALM security benchmark.
  • An open-weight model reportedly matched the strongest closed models at under a tenth of their cost, given enough turns.
  • A published limit on agent messages reportedly zeroed out tasks, so relaxing it changed rankings and favored open-weight models.
  • Different models lead different workflow stages, suggesting per-stage routing instead of choosing one model by overall score.

Why it matters for marketers: Teams building security agents can weigh routing workflow stages to different models, and should not judge models by price or a single overall score.

via arXiv cs.AIAgents & automationModels