PLUS: Anthropic’s OpenAI roast, Amazon’s $200 billion AI plan, and Meta’s AI saving lives in the trenches

Good morning, AI enthusiast.

A 40-year-old math puzzle has finally been solved, with OpenAI’s GPT-5 acting as a research partner to a UCLA mathematician. The achievement demonstrates how advanced models are becoming essential collaborators for major scientific breakthroughs.

This success marks a significant shift, showcasing AI’s growing ability to contribute to abstract reasoning, not just data analysis. As human-AI teams become the new standard in research, what other long-unsolved scientific problems could be next?

In today’s AI recap:

• GPT-5 helps solve a 40-year-old math problem
• Anthropic targets OpenAI with an ad-free pledge
• Meta’s AI powers life-saving triage robots
• The future of AI monetization: Ads vs. subs

AI Cracks a 40-Year-Old Math Puzzle

The Recap: A UCLA mathematician has finally solved a 40-year-old problem in optimization theory, using OpenAI’s GPT-5 as a research collaborator. The AI-assisted proof demonstrates how advanced models are becoming indispensable tools for scientific breakthroughs.

Unpacked:

• The breakthrough resolves a long-standing question about Nesterov’s accelerated gradient method, an optimization algorithm introduced in 1983 that is fundamental to machine learning.
• Researchers described the process as a collaborative effort, where the AI served as a partner in discovering the proof rather than just a simple calculation tool.
• This success showcases a new paradigm of human-AI collaboration, potentially accelerating the pace of discovery across other complex scientific fields.

Bottom line: This achievement shows that AI is graduating from a tool for data analysis to a genuine partner in creative and abstract reasoning. Expect to see AI-human teams tackle more of science’s greatest unsolved problems in the near future.

Anthropic Takes On OpenAI

The Recap: Anthropic is launching a major ad campaign, including Super Bowl spots, that directly targets OpenAI’s move toward an ad-supported model. In a clear counter-position, the company has committed to keeping Claude ad-free.

Unpacked:

• Anthropic argues that introducing ads creates a conflict of interest, making it difficult for an AI to act purely in the user’s best interest, especially for sensitive or complex tasks.
• The campaign is a direct response to OpenAI’s recent announcement that it will begin testing ads on the free and low-cost tiers of ChatGPT.
• Anthropic’s satirical new ads highlight the potential downsides by humorously showing Claude’s helpful advice pivoting into a predatory loan offer.

Bottom line: This marks a major divergence in how top AI labs plan to monetize their products, creating two distinct paths for the future of AI assistants. This competition will ultimately define the user experience and the level of trust we place in these powerful tools.

AI’s Two Paths: Ads vs. Subscriptions

The Recap:
The future of how we pay for AI is splitting in two. While OpenAI announced plans to test ads in ChatGPT, Anthropic immediately countered with a public commitment to keep its AI assistant, Claude, completely ad-free.

Unpacked:

• Anthropic’s commitment to Claude frames the AI as a trusted tool for deep thinking, arguing that an ad model would introduce a conflict of interest and compromise user trust on sensitive tasks.
• OpenAI is exploring an ad model to help fund broader access to its tools, promising that ads will be clearly labeled and will not influence ChatGPT’s answers or compromise user privacy.
• This divergence creates different incentives for each company, with Anthropic focused on direct user value while an ad-supported model could prioritize maximizing user engagement and time spent in-app.

Bottom line:
This decision is about more than just seeing ads; it defines the fundamental relationship between you and your AI. Users will soon face a clear choice between a free, ad-supported assistant and a paid, private workspace for their tasks and ideas.

AI in the Trenches

The Recap: A University of Pennsylvania team is deploying autonomous robots powered by Meta’s open-source AI to rapidly assess injuries in mass casualty simulations for a DARPA challenge. The system, known as PRONTO, uses models like the Segment Anything Model (SAM) to identify casualties and inform first responders.

Unpacked:

• The system uses a multi-robot team where aerial drones first locate casualties and perform initial assessments, then ground robots follow up to gather detailed vital signs and injury information.
• Team PRONTO is one of only four remaining finalists in the prestigious DARPA Triage Challenge, which aims to develop autonomous systems for large-scale emergency incidents.
• This life-saving application is powered by Meta’s open-source computer vision models, which can identify and isolate specific objects—like casualties or injuries—in complex visual data without specific prior training.

Bottom line: This project shows how open-source computer vision models are moving beyond digital applications to power real-world, life-saving robotic systems. Autonomous triage could dramatically speed up emergency response, helping first responders save more lives in critical situations.

The Shortlist

Positron closed a $230M funding round for its new inference chips, which it claims can match NVIDIA’s H100 performance while using less than a third of the power.

Amazon announced plans to increase its capital spending to $200 billion in 2026, with a heavy focus on advancing its AI, robotics, and cloud computing initiatives.

Mistral released its latest open-weight Voxtral models, featuring a real-time version for live applications with latency as low as 240ms and a highly accurate asynchronous transcription model.

Apple integrated Anthropic’s Claude Agent SDK directly into Xcode, allowing developers to use Claude’s coding capabilities natively within the IDE.