<TOP_AI_STORIES_TODAY/>

OpenAI brings frontier models and Codex to AWS

​OpenAI announced that its frontier models and Codex are now generally available on AWS, including OpenAI models on Amazon Bedrock and Codex on Amazon Bedrock, with availability across Commercial and GovCloud regions. This is a practical enterprise AI deployment move. OpenAI is not only selling better capability; it is moving frontier AI into the procurement, security, billing, governance, and cloud workflows many enterprises already use. Deployment path and procurement trust are becoming part of the product itself.

Alphabet’s $80B AI infrastructure plan

​Alphabet is turning AI capex into a balance-sheet story. According to TechCrunch, Google's parent company plans to raise $80B through a stock sale for general corporate purposes, including capital expenditures to scale AI infrastructure and global compute, at a time when demand for its AI solutions is outstripping available supply. AI is now partly a capital-allocation race: model advantage increasingly depends on whether a company can finance enough compute capacity to meet demand. For hyperscalers, the balance sheet is becoming part of the AI product stack.

Anthropic submits draft S-1 to the SEC

​Anthropic said it confidentially submitted a draft Form S-1 registration statement to the SEC for a proposed IPO. The company has not set share count, price range, or timing, and the IPO still depends on market conditions and other factors. The frontier AI race is moving closer to public-market accountability. Anthropic will not only need to prove model capability; it will face scrutiny over revenue scale, compute costs, customer concentration, and risk disclosures in a way private-market storytelling can only delay for so long.

<MICRO_SPOTLIGHT/>

✦ Balaji says AI shifts business spend toward prompting and verification

​Balaji argues that AI handles the middle of tasks better than true end-to-end work, so business spend shifts toward prompting and verification. It is an older post, but it fits today's issue: operator value moves to setup, judgment, and checking the machine's work.

✦ Sam Altman interview on human-centered AI goals

​Sam Altman says AI should not pursue goals disconnected from human needs. As agents become more autonomous, frontier-lab messaging is still trying to anchor progress around human agency, not abstract machine objectives.

✦ Aaron Levie says institutional knowledge becomes the enterprise AI moat

​Aaron Levie argues that when competitors can access similar AI models, advantage moves to internal institutional knowledge, workflows, and company-specific context. Enterprise AI differentiation is starting to look less like model access and more like context operations.

✦ SoftBank is in early talks to back an $800M Agile Robots round

​Bloomberg says SoftBank is in early talks to back an $800M round for Agile Robots. If the deal moves forward, it is another sign that AI capital is moving beyond model labs and chips into physical AI and robotics.

✦ Perplexity introduces Search as Code for AI agents

​Perplexity says Search as Code lets agents write Python that calls its search stack directly, instead of looping through function calls one by one. Search and retrieval for agents are becoming programmable infrastructure, not just another tool wrapper

<GOOD_READ/>

CEOs go bargain hunting for AI

​Axios explains why companies are routing tasks toward cheaper AI instead of defaulting everything to frontier models. Read it for the business layer behind today's issue: AI adoption is becoming a routing and cost-tier problem, not just a "which model is best" question.

Lenny Rachitsky + Benedict Evans: where AI is actually going

​Lenny points to a Benedict Evans conversation about where AI is heading and how it may affect work and life. This is the calmer strategic lens in today's package: useful after the capex, IPO, and pricing signals because it slows the reader down from news velocity to directional judgment.