Ask The PhD Community

HigherEd AI Daily

September 9 — AI Proves the Impossible; Campus Security Under Siege

Wednesday, September 9, 2026

Two landmark AI proofs in mathematics, a new writing-personalization feature in ChatGPT, and active cyberattacks on university print systems mark a consequential week for higher education and AI.

TLDR Tech — RESEARCH

OpenAI Claims to Have Solved the Navier-Stokes Millennium Problem in 88 Hours

OpenAI announced this week that its newest AI model has solved one of the seven Millennium Prize Problems: the Navier-Stokes existence and smoothness problem. The Millennium Problems, designated by the Clay Mathematics Institute in 2000 with a $1 million prize each, represent the most formidable open questions in mathematics. OpenAI's model reportedly completed the proof in just 88 hours, producing a result that claims to define conditions under which the Navier-Stokes equations completely break down.

The Navier-Stokes equations govern the motion of fluids and have underpinned physics, engineering, and meteorology for nearly two centuries. Whether smooth solutions always exist for these equations has remained unsettled since they were first formulated. OpenAI's claimed proof suggests that under certain conditions, the equations break down entirely; a result that, if verified, would have sweeping implications for computational fluid dynamics, climate modeling, and theoretical physics.

Independent verification will be essential. The mathematical community will need to scrutinize the proof rigorously; a process that could take months. Several researchers have already noted that the formal verification of AI-generated proofs at this level of complexity has no established precedent, and the announcement has already prompted significant debate in academic mathematics circles.

Why it matters for campuses

Mathematics, physics, engineering, and climate science programs will need to engage quickly with both the substance and the process of this claim. If verified, it signals that AI systems can now operate at the frontier of unsolved research problems, not merely assist with known methods. Faculty leading graduate seminars should treat this as a live case study in AI-mediated mathematical discovery; one that also raises pressing questions about authorship, attribution, and the peer review of machine-generated proofs.

Read More

TLDR AI — RESEARCH

Claude Produces the First Complete Computer-Verified Proof of Fermat's Last Theorem

Anthropic's Claude has completed the first fully computer-verified formal proof of Fermat's Last Theorem, using the Lean proof assistant over a period of 11 days. Andrew Wiles originally proved the theorem manually in 1995, in a proof spanning more than 100 pages that itself took years to fully verify by human reviewers. Claude's formalization involved 13 million lines of Lean code and the automated proof of 29,500 intermediate theorems.

The result was verified through Anthropic's Prove2Me platform. Formal verification using proof assistants like Lean ensures that each logical step is machine-checked against the rules of mathematics; a standard of rigor that human-written proofs, however brilliant, cannot match in the same automated way. The achievement demonstrates that AI can now handle the labor-intensive work of translating natural-language proofs into formally verified code, a task that historically required years of specialized human effort.

This is distinct from AI discovering a new proof; Claude did not invent the proof but rather formalized and verified an existing one with significantly greater thoroughness and speed than human teams have managed. The distinction matters for how higher education understands the role of AI in the research pipeline.

Why it matters for campuses

Mathematics and computer science departments working in formal methods, logic, or proof theory will find this result immediately relevant to their research agendas. More broadly, this development signals that the bottleneck in mathematical progress may increasingly shift from computation to the generation of genuinely novel conjectures. Graduate programs preparing students for mathematical research careers should begin discussing where human mathematical creativity fits in a landscape where AI handles formal verification efficiently and at scale.

Read More

TLDR Tech — TOOLS

ChatGPT May Soon Learn Your Writing Style Directly from Slack and Gmail

OpenAI is testing a new feature that would allow ChatGPT to study a user's writing style by analyzing examples drawn from connected apps, including Slack and Gmail. Some users have already encountered the feature in limited rollout. Once active, it would enable the model to generate text that mimics the tone, vocabulary, and structure of a specific individual; producing drafts that sound like that person rather than a generic AI assistant.

The capability is built on the same personalization infrastructure OpenAI has been developing to make ChatGPT more contextually aware of its users over time. By referencing actual communication from a user's own accounts, the model would no longer need to be told to write in a particular voice; it would already know what that means, drawing on email threads, message histories, and documents the user has written previously.

OpenAI has not officially announced the feature, which is consistent with the company's approach of testing with a subset of users before a wider launch. A screenshot of the feature interface, showing a prompt that invites users to connect apps, was published by PCMag.

Why it matters for campuses

This feature carries significant implications for academic integrity policy. If AI can convincingly replicate a student's or faculty member's writing style, existing detection tools; which often flag stylistic inconsistencies between AI-generated and human-written text; may become substantially less reliable. Institutions should anticipate this development in upcoming reviews of AI use policies, especially in programs that assess student learning through written work. Faculty and instructors responsible for academic integrity cases should be briefed on this capability before it reaches general availability.

Read More

TLDR Information Security — GOVERNANCE

Active Exploits Targeting PaperCut Print Systems in K-12 Schools and Major Universities

Threat actors are actively chaining two recently disclosed vulnerabilities in PaperCut, the widely used print management software, to steal credentials from K-12 schools and major universities across the United States and Europe. The two flaws, tracked as CVE-2026-81578 and CVE-2026-82078, allow attackers to bypass authentication entirely and then execute arbitrary code on targeted systems. Campaigns exploiting this combination have already been confirmed in the wild.

PaperCut is deployed across thousands of educational institutions for managing printers and controlling print quotas. Its presence on campus networks, often connected to systems that handle student and employee authentication, makes it a high-value target. The chained exploit allows attackers to move from print management infrastructure into broader institutional networks, accessing stored credentials that may grant access to learning management systems, email, and administrative databases.

Security researchers advise institutions to immediately restrict PaperCut servers from external internet exposure and to monitor for unusual command-line activity originating from the PaperCut process. Execution of cmd.exe or powershell.exe, and commands such as whoami or tasklist with pc-app.exe as the parent process, should be treated as indicators of compromise. Patches are available and should be applied without delay.

Why it matters for campuses

Any institution running PaperCut should treat this as an active threat requiring immediate response from IT security teams, not a scheduled patch cycle. Chief Information Security Officers and campus IT directors should verify patch status, restrict network exposure, and audit authentication systems that may have been accessed through PaperCut-connected infrastructure. The fact that K-12 and university networks are specifically named in active campaigns means this is not a hypothetical risk but an ongoing one.

Read More

Tool of the Day

ChatGPT Images 2.5

OpenAI released ChatGPT Images 2.5 this week, a substantially upgraded image generation model available to all ChatGPT, ChatGPT Work, and Codex users across desktop, mobile, and web. The new model produces sharper details and more precise edits than its predecessor, and generation latency has been reduced by up to 50 percent compared with Images 2.0. A new Sketch feature allows users to draw directly in the app as a visual reference for the model.

For educators, this release meaningfully lowers the barrier to creating polished visual materials without graphic design expertise. Syllabi illustrations, presentation diagrams, concept visualizations for abstract topics, and custom images for instructional handouts are all within reach using plain-language prompts in a tool many faculty already have access to.

Try it: Open ChatGPT and use the image generation panel to create a visual analogy for a concept students consistently find difficult in your course; for example, a diagram illustrating the difference between correlation and causation, or a stylized image representing a key event from your syllabus.

Visit ChatGPT Images 2.5

Have a great learning day!

Dr. Ali Green

Sources for This Edition

TLDR Tech (tldrnewsletter.com)
TLDR AI (tldrnewsletter.com)
TLDR Information Security (tldrnewsletter.com)
PCMag (pcmag.com)
Anthropic Research (anthropic.com)
The Hacker News (thehackernews.com)

askthephd.com
 | 
askthephd.substack.com
 | 
Unsubscribe

HigherEd AI Daily; Curated by Dr. Ali Green