OpenAI may shelve ChatGPT Agent just months after launch. Users dropped from four million to under one million, plagued by technical issues and unclear purpose: many didn’t know what to use it for or that it even existed. The branding didn’t help either, suggesting only this mode was agentic when ChatGPT already had agent capabilities.
Google Deepmind has made Project Genie publicly available. The experimental prototype, based on the Genie 3 world model shown in August, is now accessible to Google AI Ultra subscribers in the US who are 18 or older.
The web app lets users create interactive worlds using text or images and explore them in real time. The system generates the environment as you move through it. Project Genie offers three main features: World Sketching for creating worlds with Nano Banana Pro and Gemini, World Exploration for moving through them, and World Remixing for changing existing worlds.
Google says the prototype still has issues: worlds don't always look realistic, characters sometimes respond slowly, and sessions are limited to 60 seconds. Some features announced in August, like promptable events, are still missing. Google plans to expand to other countries later.
The long-term goal of such world models is to serve as training environments for AI agents, allowing them to learn from simulated experiences instead of relying solely on pre-collected data.
OpenAI researcher Kevin Weil pushes back on reports that the company plans to claim a share of discoveries made by individual users, entrepreneurs, or scientists. The clarification follows a blog post by CFO Sarah Friar outlining plans for IP licensing agreements and outcome-based pricing that would let OpenAI share in the value its tools help create.
Licensing, IP-based agreements, and outcome-based pricing will share in the value created.
Sarah Weil, via OpenAI
Weil clarified on X that Friar was referring to interest OpenAI has heard from large organizations in licensing or IP-based partnerships. The company is open to exploring creative ways to partner and align incentives, but "that's not something we're doing today." If it happens in the future, it would be a bespoke agreement with a company, "not something that would impact individual users," Weil says.
Microsoft posts record cloud revenue but the stock is down double digits. Investors question whether billions in AI spending will pay off, especially with nearly half the cloud backlog coming from one customer: OpenAI.
OpenAI's latest funding round might hit peak circularity. According to The Information, the AI company is in talks with Nvidia, Microsoft, and Amazon about investments totaling up to $60 billion. Nvidia could put in as much as $30 billion, Amazon more than $10 billion—possibly even north of $20 billion—and Microsoft less than $10 billion. On top of that, existing investor SoftBank could contribute up to $30 billion. If these deals go through, the funding round could reach the previously rumored $100 billion mark at a valuation of around $730 billion.
Critics will likely point out how circular these deals really are. Several potential investors, including Microsoft and Amazon, also sell servers and cloud services to OpenAI. That means a chunk of the investment money flows right back to the investors themselves. These arrangements keep the AI hype machine running without the actual financial benefits of generative AI showing up in what end users pay.
AI research institute Allen AI has released SERA, a family of open-source coding agents designed for easy adaptation to private code bases. The top model, SERA-32B, solves up to 54.2 percent of problems in the SWE-Bench-Test Verified coding benchmark (64K context), outperforming comparable open-source models.
According to AI2, training takes just 40 GPU days and costs between $400 to match previous open-source results and $12,000 for performance on par with leading industry models. This makes training on proprietary code data realistic even for small teams. SERA uses a simplified training method called "Soft-verified Generation" that doesn't require perfectly correct code examples. Technical details can be found in the blog.
The models work with Claude Code and can be launched with just two lines of code, according to Allen AI. All models, code, and instructions are available on Hugging Face under the Apache 2.0 license.
Just last October, Andrej Karpathy dismissed AI agents: “They just don’t work.” Now he says 80 percent of his coding is agent-based and calls it the “biggest change to my basic coding workflow in ~2 decades.” A typically measured voice is joining the agent coding hype, but with some warnings attached.
OpenAI is charging around $60 per 1,000 impressions for its initial ChatGPT ads, far above typical online advertising rates in the low single digits and closer to what advertisers pay for premium TV spots like NFL games, according to The Information. The ads show up below ChatGPT responses in the free and lower-cost "Go" tiers.
OpenAI is also reportedly charging per impression rather than per click. Advertisers typically prefer click-based billing because it's easier to measure results. The decision to go with impressions likely reflects how AI chatbot users behave differently than traditional search users: they click on external links far less often. Perplexity uses the same approach, also charging per 1,000 impressions.
The move toward advertising—at premium prices and in a format that's less appealing to advertisers—suggests OpenAI needs to ramp up revenue quickly to justify its high valuation to investors. Sam Altman previously called ChatGPT advertising a last resort and a potential dystopia.
