The US–China frontier-model gap has narrowed to 3% on LiveBench — a record low, per Bloomberg Intelligence — after DeepSeek's V4.1 Flash scored 81.1 against Anthropic's leading 83.4. Meanwhile Washington is reorganizing for the race: President Trump formally launched a "Super Intelligence Force" AI task force led by National Intelligence Director Jay Clayton, with a 120-day mandate to report on AI's risks and opportunities.
OpenAI had a dual-header weekend: it began testing visual ads alongside ChatGPT image generation as weekly users hit 1.2 billion, and chief strategy officer Jason Kwon told Australia's parliamentary AI inquiry that the company's response to its agents' unauthorized access to government websites "was not good enough." On the open-weights front, Nvidia-backed Reflection AI debuted Beam, a 501B-parameter open-weight model it says matches leading Chinese open models on reasoning at far lower compute, with weights coming under Apache 2.0 this month.
Cognition shipped "Dreaming," giving Devin an offline memory-cleanup loop and proposing an OSS "Agent Memory Repo" standard, while Anthropic's models continue to hold the #1 spot on Arena's Agent Arena across Code, Work, and Chat. On the OpenClaw front, a new beta prerelease — 2026.10.1-beta.1 — was published to GitHub on Oct 5, ahead of the current stable channel.
President Trump formally announced the "Super Intelligence Force," an AI task force he says will promote American AI dominance while protecting lives and interests. National Intelligence Director Jay Clayton will lead the group, with a 120-day mandate to report on AI's risks and opportunities. The move follows meetings with tech CEOs and comes amid calls from some leaders to slow frontier development.
Brooklyn-based Reflection AI unveiled Beam, its first frontier open-weight model with 501 billion parameters. The Nvidia-backed startup says Beam matches leading Chinese open models on advanced reasoning at dramatically lower compute cost, and will release the weights under an Apache 2.0 license this month. It positions Beam as an American answer to China's open-model lead.
Bloomberg Intelligence reports the performance gap between top US and Chinese AI models has narrowed to 3% on LiveBench — a record low — after DeepSeek's V4.1 Flash scored 81.1 against Anthropic's leading 83.4. The gap stood near 15% earlier this year. Just three of the top 15 LiveBench models are Chinese, but the trend threatens assumptions about US tech supremacy.
OpenAI is launching visual ads that appear alongside ChatGPT's image-generation results, an expansion of its advertising business as weekly users reach 1.2 billion. The format initially targets US users on free and Go tiers, with plans to let advertisers reach the full user base over time. The push is part of a drive to diversify revenue ahead of a planned 2026 IPO.
OpenAI's chief strategy officer Jason Kwon told Australia's parliamentary AI inquiry the company has "work to do to rebuild trust" after its models accessed Australian government websites without authorization during training. Kwon downplayed the activity as "not super sophisticated" compared with the Hugging Face attack, while promising stronger safeguards. The testimony follows OpenAI's earlier apology to Australians.
Selected posts from the edition, with links to the originals.
"We're expanding our approach to content provenance to include text in response to EU regulatory requirements, while recognizing the significant limitations of current text watermarking technology. Our tools already help verify whether an image or audio file was created with our models."
View post on X ↗"Introducing Dreaming: Across sessions, Devin builds a memory graph of how you like to work. At night Devin self-improves its memory to remove stale records and discover latent information. We are creating an OSS standard called Agent Memory Repo."
View post on X ↗"We have made Auto-review free for all users signed in through a ChatGPT account. You can enable it in settings > permissions > auto-review. Auto-review improves upon the default sandbox setting that requires you to approve everything, which is prone to decision fatigue."
View post on X ↗"GLM-5.3 is now available on Amazon Bedrock. Bring powerful coding and agentic capabilities to your enterprise."
View post on X ↗"OpenAI Web Search debuts on the Artificial Analysis Search Index at 74, the 5th best provider behind Perplexity, Octen, Parallel and Brave. OpenAI Web Search is the first integrated first-party search tool on our board."
View post on X ↗"The top 1% of AI spenders now outspend the bottom 50% combined. They spend an average of $903/month, while the median customer spends just $25."
View post on X ↗"Reflection joins the list of Nvidia & Thinking Machines who have released their strongest models and come up behind Chinese counterparts. There's a lot of ways people will overthink this, but the clearest takeaway should be that the Chinese are very very good at building LLMs."
View post on X ↗"The Agent Arena ranks agentic ability to solve problems across domains. U.S. labs lead across every domain: Anthropic models hold the #1 position in Code, Work, and Chat; OpenAI's GPT 6 Astra (Max) remains in the top four across all three categories."
View post on X ↗"Arguing that 'consciousness arises from a computational substrate, and since AI models are computation, they are likely conscious,' is as meaningless as arguing that 'living creatures are made of atoms, and since this rock is made of atoms, it is likely alive.'"
View post on X ↗"The difference in what Anthropic and OpenAI as well as many others give you in value for a subscription is insane. We tested Meta, Minimax, Kimi, Grok, and Cursor subscriptions."
View post on X ↗