People often ask how I keep up with AI and decide what is worth following. This issue skips the usual essay and shares the publications, shows, and people that make up my current rotation.
The point is not to read or watch everything every day. Each source has a job. Daily shows and newsletters help discover stories. Established reporting and original sources verify claims and add context. Longer interviews reveal how people actually apply the technology. X provides early notice of releases and technical debates, but a post there is not confirmation by itself.
How I Stay Informed About AI - The main feature. The sources I return to most include The AI Daily Brief for major model, business, policy, and market stories; How I AI for practical demonstrations; AI & I for founder and builder workflows; Lenny's Podcast for long operating conversations; AI News & Strategy Daily for business consequences; and Peter Yang for product, agent, and team-workflow guides.
Broad newsletters such as The Rundown, Superhuman, AI Secret, and Ben's Bites are useful discovery scans, though overlapping coverage does not count as independent confirmation. Import AI brings a research and policy lens, Postcards From the Edge connects AI to strategy and organizational change, and IBM Think offers enterprise explainers across AI, security, infrastructure, and operations.
For reporting and analysis, the Financial Times, Wall Street Journal, and Economist put announcements into business, economic, and policy context. MIT Technology Review is particularly useful for research, science, policy, and technical limits. VentureBeat often surfaces enterprise implementation details, while WIRED and the New York Times add reporting on the people and institutions around the technology. BCG publications provide an enterprise operating-model perspective, with the evidence weighed separately from the framework.
The people I follow on X also have distinct roles. Andrej Karpathy, Erik Brynjolfsson, and Jack Clark cover research and AI's effect on work. Builders such as Boris Cherny, Thariq Shihipar, Tibo Sottiaux, and Peter Steinberger provide early product and technical signals. Claire Vo, Peter Yang, Dan Shipper, and Tibo Maker share workflows and independent experiments. Official lab and project accounts remain useful for releases, but those announcements still need verification and context.
Also in this issue:
- The Brief - Cisco's 90,000 employee agents turn ownership and performance into management questions; AI assistants are becoming the front door to company software; company context matters more than model quality alone; and heavy enterprise users are building reusable instructions and system connections.
- Patterns & Signals - Agent inventories and retirement criteria are becoming necessary at scale, while Salesforce and Anthropic are moving CRM actions and managed agent infrastructure into Claude.
- Signals to Watch - Nvidia is licensing Poolside technology, OpenAI published the Hugging Face incident postmortem, BCG reported a large gain from mapping system dependencies, private equity is embedding AI engineers, and EU AI Act enforcement has begun requesting information.
- Meanwhile... - KU Leuven researchers are using 1.4 million passes and 60,000 throw-ins to test counterintuitive soccer tactics that coaches have traditionally judged by feel.
- What I'm Consuming - Building useful AI evaluations, the widening gap among enterprise adopters, systems of record in agent workflows, the loss of junior learning pipelines, and why benchmarks must match real workloads.
- Term of the Week - Open-weight models, whose learned weights can be downloaded and run on compatible infrastructure even when training data and licensing remain restricted.
- After Hours - Georges Simenon's Act of Passion, a concise psychological confession about love, possession, self-justification, and a life shaped by other people.
The useful habit is to separate discovery from verification. Let fast sources surface what may matter, then return to original announcements, established reporting, research, and operating evidence before deciding what to believe or apply.