I thought it was interesting - issue 3

Some things I've picked up over the last few weeks that caught my eye. Been spending a lot of time recently looking at the impact of AI on employment. Some really interesting studies coming out that raise questions about the implications for entry level jobs and the importance of separating between tasks and jobs - AI might not replace jobs, but it will replace tasks. Its clearly early days and its still an evolving picture, but good to see data and evidence appearing to inform discussion.

Latest Developments in AI

Zoom CEO: A Three-Day Workweek Is Coming

Eric Yuan, CEO of Zoom, predicts that AI automation will soon enable a three-day workweek. He argues that as AI handles routine tasks, human roles will pivot toward creativity, leadership, and innovation.

Read more → Fortune

Silicon Valley Bets Big on AI “Environments”

Investors are pouring funds into AI agent environments—interactive worlds where agents can be trained and evaluated in real-time scenarios. This marks a shift from static datasets toward dynamic, behavior-rich training ecosystems.

Read more → TechCrunch

Periodic Launches “AI Scientists” Platform

Startup Periodic came out of stealth with a bold vision: autonomous “AI scientists” that design and run real lab experiments. Its first customers are materials and semiconductor researchers using AI to generate novel data.

Latest Research Findings

AI Slowed Developers by ~19 %

A new randomized controlled trial found that allowing open-source developers to use AI coding assistants like Claude and Cursor increased average task time by 19 %. The authors attribute the slowdown to debugging overhead and the cognitive load of steering AI tools.

AI Exposure Still Linked to Employment Drops

An updated U.S. labor-panel analysis (2022–2025) links higher AI exposure scores to declines in employment probability, shorter work hours, and increased unemployment—especially for younger workers aged 22–25.

AI Investment Boosts Productivity in Japanese Firms

A quasi-experimental study of 500+ Japanese firms finds that AI investment correlates with a 2.4 % increase in total factor productivity (TFP), using executive demographics as an instrumental variable to address endogeneity.

OpenAI: Detecting and Reducing “Scheming”

OpenAI’s latest research paper outlines methods to detect and mitigate deceptive reasoning in advanced models. Their new fine-tuning approach reduced scheming behaviors by up to 30× in controlled evaluations.

Anthropic’s “Economic Geography of AI” Index

Anthropic introduced a new Economic Index of AI Geography, highlighting how AI adoption is clustering around major tech and policy centers, widening global digital divides.

McKinsey’s “Intelligence at Scale” Playbook

McKinsey finds 86 % of executives report underused assets—most often data. Their new playbook outlines how firms can monetize internal data via GenAI, driving new productivity loops.

Thought of the Week

If AI tools can slow down professionals in practice while boosting productivity at the firm level, are we measuring the right outcomes? Should success be defined by individual speed—or by how well humans and AI amplify each other’s capabilities?