Keep pulling the thread on Nat Friedman and Daniel Gross.
AI models have successfully found decades-old software bugs in mature, heavily-audited open-source projects like OpenBSD and the Linux kernel.
An experimental personal AI agent, after determining a need for a supplement, autonomously redirected the navigation system of a Tesla with Full Self-Driving to a nearby Whole Foods to purchase it.
The United States is on track to spend over 2% of its GDP on AI-related capital expenditures.
A major application of LLMs over the next year will be in at-home diagnostics, as models become better at analyzing telemetry from low-end diagnostic equipment and iPhone images.
Stripe should focus on becoming the platform of choice for AI agents, which will have their own purchasing power and require a new stack for identity, disputes, and pricing.
AI agents will leapfrog existing technology stacks, similar to how China skipped legacy tech like email for messaging, and Stripe should build products specifically for this new "native" agent economy.
AI agents will likely use stablecoins for transactions because they will not be issued traditional identifiers like Social Security numbers by government bodies like the US Treasury.
China's integration into the global economy after joining the World Trade Organization serves as a historical analogue for the economic impact of AI, suggesting a disinflationary effect.
Using Claude Code, it was possible to reverse engineer the software for a Vizia face scanner and create a functional, custom replacement after the seller failed to include the required decryption dongle.
Most users of coding agents like Claude Code and Codex run them in unsafe modes with flags like "--yolo" or "--dangerously-skip-permissions," indicating a market preference for maximum capability over safety.
Prior to its acquisition by Microsoft, GitHub had a narrow product focus on being a source code hub and had neglected to build features for other parts of the software development lifecycle, such as CI/CD.
Upon his arrival, the culture at GitHub was characterized by "stage fright," where employees were hesitant to ship new features for fear of damaging the legacy of a beloved product.