OpenAI's Math Win Comes With Allegations

OpenAI announced Tuesday that one of its internal models solved the Navier-Stokes problem, a long-standing mathematical puzzle about how fluids move, according to The Verge. The feat took 88 hours, used a swarm of roughly 10,000 AI agents working in parallel, and comes with a $1 million prize attached — the problem has stumped human researchers for nearly a century.

The result itself is significant. What has rattled the mathematics community is how OpenAI got there.

One day before OpenAI's announcement, New York University professor Tristan Buckmaster and Levent Alpöge — an Anthropic researcher working independently — published findings on a closely related problem. Buckmaster says he reached out to OpenAI after learning the company knew about his research. The exchange, he says, turned hostile. An OpenAI researcher allegedly told him that going public would "ruin his career." OpenAI also reportedly urged him to drop Alpöge as a co-author and credit OpenAI's model instead.

Buckmaster asked whether OpenAI had accessed his activity on Codex, OpenAI's own coding tool he had used during his research. OpenAI denied using any specific user data, but could not fully rule out indirect influence: "while unlikely, we cannot rule out that de-identified data derived from their usage of our products helped improve our models," the company said.

Allegations remain unproven and the full timeline is still unclear. What is clear is that several mathematicians believe OpenAI violated academic norms that researchers rely on to share work openly.

Watch for whether OpenAI publishes its proof for independent peer review. That step would at least put the mathematics itself on firmer ground.

GPT-Live-1 is a voice model from OpenAI that listens and speaks at the same time, rather than waiting for you to finish before it responds. It was first built into ChatGPT; it is now available to developers building their own voice-powered apps and services, according to OpenAI.

Traditional voice systems stitch three separate tools together: speech-to-text, a reasoning model, then text-to-speech. Every handoff adds delay and creates new ways for the conversation to fall apart. GPT-Live-1 handles all of that inside a single model, which means it can react to interruptions in real time instead of losing track of what was being said.

According to OpenAI's own evaluations, it outperforms its predecessor GPT-Realtime-2.1 on Full Duplex Bench — a test for how naturally a voice agent handles conversation flow — by 30 percentage points, though this evaluation was not independently verified.

It is aimed at developers, not general consumers. Pricing is not specified in the announcement.

Honest read: Worth watching if you build anything voice-related — real-time interruption handling is a genuine gap in current voice tools. If you are not a developer, this one is not for you yet.

What it is: A fully assembled image-generation workspace built on Gradio's new Workflow system, replicating most of the features of AUTOMATIC1111's stable-diffusion-webui — a popular open-source tool for generating images with Stable Diffusion models.

What's new: Previous Gradio tools required writing code to connect models together. Workflow1111 replaces that with a visual canvas where models, functions, and data sources are laid out as nodes you connect by drawing edges between them.

How it works: The canvas contains eleven media pipelines built from seventy-three individual nodes. Each node wraps one operation — a Python function (fn), a model call, another Gradio Space, or a dataset row. Connecting a node's output port to another node's input port defines how data flows. Because nodes with shared inputs run simultaneously, tasks like image captioning and label classification happen in parallel rather than waiting in sequence.

What it can't do: Running the pipelines uses your own Hugging Face account quota, so heavy use will hit rate limits. The post does not address performance on large or high-resolution generation tasks, and the canvas approach may be harder to debug than code when something breaks unexpectedly.

Who should care: Anyone building or experimenting with image generation workflows who wants to skip the setup overhead of wiring models together manually.

Try it / read it: Workflow1111 on Hugging Face

IBM has released Granite Time Series PatchTST-FM-r2, a forecasting model that predicts future values in any numerical time series — sales figures, energy loads, website traffic, sensor readings — without needing to be trained on your specific data first. That approach is called zero-shot forecasting (generating predictions from a pre-trained model, with no fine-tuning required on new data).

On the GIFT-Eval leaderboard, a benchmark for time series forecasting run by independent researchers, the model ranks second overall among comparable zero-shot models. It is first among models carrying a permissive commercial licence — specifically Apache 2.0 (a licence that lets you use, modify, and deploy the model in commercial products without royalty fees).

The model weights, code, and benchmark reproduction steps are all available on Hugging Face today: Granite Time Series PatchTST-FM-r2.

Worth knowing: The model requires approximately 385 million parameters to run — check that your available hardware can handle that before committing to it for production use.

Microsoft has agreed to ten AI privacy principles with the American Federation of Teachers and its New York City affiliate, according to The Verge. The commitments include not training AI models on student or teacher data, limiting data collection, banning AI companions, and requiring human review for high-risk decisions. School districts can attach these terms to existing Microsoft contracts from November onward. The agreement follows one-year bans on student-facing AI tools in both New York City and Los Angeles schools.