In a lecture at Stanford's CS146S course, Lee Robinson argued that every new product must include agent capabilities or risk becoming headless data for other agents. He outlined four eras of developer-AI collaboration, envisioned agents moving toward an 'always-on' form, discussed cost-control principles such as thin-client fixed-server architecture, main/sub-agent coordination, prompt caching, and 'deleting products,' and predicted agents will be able to work continuously for weeks or months within the next six to twelve months.
On Sept. 21, researchers from Stanford University and King's College London published a paper based on 1.25 billion job postings and 154 million employment records from 41 countries (Jan. 2021–Mar. 2026), finding that AI-adopting firms added more jobs but tilted toward senior staff, with senior roles up 6.7% over five years, entry-level jobs down 3%, the junior share down 1.9 percentage points, and computer/math job share up 0.8 percentage points.