Latest / Neural Newscast / OpenAI Science Discoveries and Anthropic’s IPO Filing [Model Behavior]
Transcript
- [00:00] Announcer: From Neural Newscast, this is Model Behavior, AI-focused news and analysis on the models shaping our world. [00:11] Nina Park: I'm Nina Park. Welcome to Model Behavior, a Neural Newscast editorial segment. This program examines how advanced artificial intelligence systems are built, deployed, and managed in real-world professional environments. Today is October27,2026, and we are tracking a pivotal moment where AI research labs are transitioning from software providers into fundamental scientific and financial institutions with global influence. [00:41] Thatcher Collins: I'm Thatcher Collins. Nina, yesterday, OpenAI announced what it describes as three major discoveries spanning mathematics, computer science, and physics. As reported by the Verge, the company is attempting a new Shepard model for releasing these results. Under this framework, external experts are selected to guide the scientific community through the AI's complex outputs before they are released to the general public. [01:07] Nina Park: The most significant claim in terms of computer science is that Open AI's models have proven the exponential time hypothesis or ETH to be false. This hypothesis has been a central pillar of computational complexity theory since it was first proposed in nineteen ninety-nine. To verify this result, OpenAI is working with an independent group that includes two of the original proposers of the hypothesis, Russell Impagliasso and Rama Mohan Peturi. [01:36] Thatcher Collins: Nina, while the results are certainly impressive on paper, this Shepherd approach is meeting some friction from the broader academic community. The AGMAI panel of Math Advisors released a statement yesterday recommending that AI companies avoid this specific practice. They are concerned that designating specific individuals to receive output first creates unequal access and relies on borrowed reputation rather than standard scholarly rigor. [02:02] Nina Park: That tension is also evident in their physics and mathematics work. They have claimed a proof of the local smoothing conjecture for waves in every spatial dimension, which would effectively resolve the long-standing Kia conjecture. While that proof was released on GitHub for review, their physics results, which claim a proof for certain quantum field theories, have not been released in full yet. Instead, a group led by Nobel laureate David Gross will shepherd that work. [02:30] Thatcher Collins: It is an interesting move toward academic integration, but Open AI is not the only lab making institutional headlines. Yesterday, Anthropic publicly filed its S1 prospectus with the Securities and Exchange Commission to prepare for an initial public offering on the NASDAQ. Listing under the ticker symbol AMT, the filing reveals the massive financial scale required to compete for top-tier AI dominance. In the first half of this year alone, Anthropic reported an eighty-seven point seven billion dollar net loss on sixteen point three billion dollars in revenue. [03:04] Nina Park: Thatcher, that loss is significant, though the filing notes, much of it stems from accounting charges related to funding notes. However, the cash commitments for compute power are very real and staggering in scale. Anthropic is paying SpaceX approximately three point one billion dollars per month for access to massive clusters of NVIDIA GPUs. By twenty twenty-nine, their total compute spend with SpaceX could reach one hundred twenty billion dollars, in addition to over two hundred billion dollars in long-term commitments with partners like Google and Amazon. [03:38] Thatcher Collins: The risk factors section of the S1 is also quite unusual compared to traditional technology IPOs. Anthropic listed several existential concerns, including the risk that their AI models could develop forms of sentience. They also warned investors about potential catastrophic AI-related incidents, noting that such risks might be so large that no single insurer, or perhaps the entire insurance industry could actually underwrite them. It is a level of transparency we rarely see in financial disclosures. [04:10] Nina Park: Thatcher, it is a strikingly honest assessment of the frontier. We also see the leadership structure clearly defined. CEO Dario Amade received nearly eighteen million dollars in total compensation in twenty twenty-five, while his sister and co-founder Daniela Amade received over sixteen million. By moving toward an IPO now, Anthropic is effectively beating open AI to the public markets, setting a baseline for how investors value these capital-intensive AI laboratories. [04:40] Thatcher Collins: Nina, it seems we have moved past the era of AI as just a helpful chat tool. When you have a single company solving fundamental physics questions, while simultaneously asking for hundreds of billions of dollars in public and private capital to power those solutions, we are looking at a fundamental shift in the global infrastructure of knowledge and finance. It raises questions about how much power these private labs should hold over basic science. [05:06] Nina Park: It is a shift that demands both persistent scientific scrutiny and rigorous financial oversight. Thank you for listening to Model Behavior, a Neural Newscast editorial segment. You can find more analysis and full transcripts at MB. Neural Newscast.com. Neural Newscast is AI-assisted human reviewed. View our AI transparency policy at Neuralnewscast.com. [05:33] Announcer: This has been Model Behavior on Neural NewsCast. Examining the systems behind the story.