BIG TECHAWAYS

OpenAI is starting to touch the silicon layer

• ​Reuters reports that the general manager of OpenAI Korea said the relationship with Samsung includes joint production and research into next-generation chips. The companies have not disclosed specifications, scale, or a specific chip product.
• The report points to an AI lab moving deeper into supply-chain coordination. As compute becomes an operating constraint, advantage depends not only on the model but also on coordination with manufacturers, infrastructure partners, and the silicon underneath.
• The public information is not enough to conclude that OpenAI has its own chip, has secured capacity, or is preparing a commercial product. Reuters is the source for the relationship; the scope of the work remains undisclosed.

Meta is teaching Muse when to ask again

• ​Meta describes Muse as an always-on personal agent with an isolated VM for each user. The system includes Sentinel action checks, limits on the credentials an agent can see, and confirmation before sensitive actions. The initial rollout is limited to users in the U.S.
• Operational boundaries are not an afterthought. Where an agent runs, what it can see, and when it must ask are all part of the product design. For agents that touch email, calendars, and browsers, recovery, access scope, and data handling may matter as much as the number of tasks completed.
• Meta is the source of the architecture and rollout details, and they are not the result of an independent security audit. The direction is clear: permissions and observability are becoming part of the agent experience.

An AI proof still has to pass through Lean

• ​OpenAI says a group of agents using a next-generation model produced an analytical proof and a Lean formalization for a result involving a forced 3D Navier-Stokes flow singularity.
• The important part is the pipeline. The agents did not only generate a solution; they also moved it into a formal-verification system. Combining multiple agents, specialized tools, and a repeatable checking layer opens a new path for automating parts of research.
• This remains an OpenAI-reported result. OpenAI has not provided public independent confirmation, so it cannot yet be treated as a solution validated by the mathematical community. The current significance is the organization of the proof-generation and checking process, not a settled claim that the community has accepted.

PRESENTED BY NIBBLES

You work a 9-to-5 so your pet can live like royalty. Get rewarded for it.

Your pet lives like a king on your salary: organic treats, memory-foam beds, and endless toy drops. Isn't it time their high-maintenance lifestyle worked for you? Meet Nibbles, the first credit card engineered specifically for pet parents. Earn 3x rewards on pet & vet expenses, from organic pet food to grooming, and get 1x rewards everywhere Mastercard is accepted. You can keep spoiling your highness without breaking the bank. Get the no annual fee card that actually rewards you for spoiling your pet.

​Apply now​

Sponsored by Nibbles. Nibbles is not a bank. The Nibbles Card is issued by Lead Bank. Other fees and T&Cs apply. Rewards & pet insurance offered by Nibbles, not Lead Bank. Base pet insurance policy paid for one eligible pet by Nibbles. See nibbles.com for details. Nibbles does not operate in CT, NY, MN, WA, CO, HI, MA, or ME.

BIG THINK

SaaS is getting sold off, but disruption is not arriving fast

• ​Asa Fitch of The Wall Street Journal describes a less dramatic picture than the idea that AI will erase SaaS. Companies such as Salesforce and Workday have seen their shares hit, yet still show financial strength, while disruption is arriving more slowly than many feared.
• That separates two questions that are often bundled together: how much cheaper and easier AI can make software creation, and whether companies that own enterprise systems will lose revenue immediately. Lower code-creation costs do not automatically remove the cost of maintenance, security, data, integration, and operational responsibility.
• This picture does not prove incumbents are safe. It shows that perceived disruption and enterprise economics can move at different speeds.

AROUND THE AI WORLD

• OpenAI splits Images 2.5 into faster and more controlled paths: OpenAI introduced Sketch and templates, along with Flare for speed and Sunburst for greater control. Creative workflows are separating throughput from precision instead of using one default model for every asset.
• Cognition raises more than $2 billion at a $48 billion valuation: Cognition says its Series E raised more than $2 billion at a $48 billion valuation. It is a major platform bet on coding agents, not proof that adoption is durable.
• Qualcomm and Amazon collaborate on inference silicon and optical connectivity: Qualcomm says the collaboration includes customized silicon for AI inference and optical connectivity of up to 1.6T. The infrastructure race now spans compute, networking, and power efficiency.
• AlphaGenome Atlas turns predictions into a research surface: Google DeepMind says AlphaGenome Atlas precomputes predictions about the molecular effects of roughly 9 billion single-letter human genetic variants. Its value is the query layer for researchers, not clinical use.