What I am seeing in current builds, frontier AI, and software calls. Written so an operator can decide what to do next.
How we apply thisNewsletter
A weekly digest of new field notes on AI, software delivery, and governance. No spam. Unsubscribe anytime.
Browse by focus
The three piles I write in most. Counts are matching notes in the archive.
AI
Perplexity Personal Computer for Mac dropped yesterday. The AI runs 24/7 on your Mac, controls your files and apps natively, and can be managed from an iPhone. This is not a chatbot with a new coat of paint. This is a fundamental shift in what AI does for you.
Anthropic
On April 16, 2026, Anthropic launched Claude Opus 4.7 and confirmed it has crossed $30 billion in annual recurring revenue, surpassing OpenAI. The company that spent two years being called "the safety-focused alternative" is now the commercial leader of the AI industry.
Anthropic
A leaked report on April 15 revealed Anthropic is launching Claude Opus 4.7 with a design tool that builds websites and presentations from natural language prompts. Adobe, Figma, Wix, and GoDaddy all dropped over 2% in a single session. This is the second time in months that one Anthropic announcement moved markets outside the AI industry.
AI Regulation
xAI filed suit in federal court to kill Colorado's SB 24-205, the nation's first AI anti-discrimination law. If the company wins, state AI regulation dies before it starts. If Colorado wins, expect 20 states to follow. The stakes are that simple.
OpenAI
OpenAI announced a multi-tier cybersecurity access program on April 14, 2026, pairing GPT-5.4-Cyber with Aardvark agentic security researcher and a new Frontier Risk Council. This is not just a product update. It is OpenAI matching Anthropic's controlled rollout strategy move-for-move. The AI security race is now a direct two-horse competition.
AI
UK regulators are in emergency talks with banks and the NCSC over Anthropic's Claude Mythos. The US Treasury already briefed Wall Street. This isn't about model capabilities anymore. It's about whether frontier AI breaks the financial system's threat model.
Anthropic
Reuters reports Anthropic is exploring custom AI chip design. The plans are early and nothing is committed, but the signal matters more than the silicon. When a company growing this fast starts eyeing chip design, the compute bottleneck is real.
AI
OpenAI acquired TBPN, a tech talk show with a small but influential audience of founders and operators. The deal has nothing to do with content. It's about controlling where the AI narrative forms before it hardens into consensus.
AI
Meta signed a $21 billion cloud capacity deal with CoreWeave, extending through 2032. The deal includes early deployments of NVIDIA's Vera Rubin platform and expands a $14 billion agreement from last year. These numbers stopped surprising me, and that's the part that worries me.
OpenAI
An Axios report appeared to link OpenAI's newest model to a restricted cybersecurity rollout. OpenAI says the two are unrelated. The real story? Reporting around frontier AI capabilities is now part of the risk surface, and the gated-access trend Anthropic kicked off is still accelerating — just with worse signal-to-noise than anyone expected.
AI
Three new in-house models, MAI-Transcribe-1, MAI-Voice-1, and MAI-Image-2, are live in Microsoft Foundry. The models themselves are fine. The real story is what they say about Microsoft's relationship with OpenAI.
AI
Chinese state-sponsored hackers used Anthropic's AI to run 80-90% of a cyber espionage operation against 30 companies and government agencies. The attacks hit thousands of requests per second. Human oversight: four to six decision points. This is not a future problem. This is now.
Anthropic
Anthropic just previewed Mythos through Project Glasswing, a tightly controlled cybersecurity program that pairs a frontier model with major defensive partners. This is bigger than a product launch. It is a bid to write the rules of AI-powered cyber defense before governments can.
AI
As of today, Claude Pro and Max subscribers can no longer use their plans to power tools like OpenClaw. The reason is "capacity." The timing is not coincidental. Here's what actually happened and why it matters.
AI
Google Research dropped a compression algorithm that makes AI models 6x smaller and 8x faster with zero accuracy loss. No retraining required. Memory chip stocks are cratering, and Cloudflare's CEO is calling it "Google's DeepSeek moment."
AI Safety
A data leak exposed Claude Mythos, Anthropic's unreleased AI model that the company admits is "far ahead of any other AI model in cyber capabilities." The irony of a cybersecurity-focused model being revealed through a security failure writes itself.
