weekend ai reads for 2025-05-30

“We don’t circulate this too much in the AI community – not just our models but all models – tend to do better if you threaten them … with physical violence,” he said in an interview last week on All-In-Live Miami.

All self-modifications and evaluations occur within secure, sandboxed environments, under human supervision and with strict limits on access to the web. The DGM archive also provides a transparent, traceable lineage of every change, and our reviews of the changes that occurred in our experiments show modifications focused solely on improving coding capabilities.

We live in the era of the symbolic executive, when “being good at stuff” matters far less than the appearance of doing stuff, where "what's useful" is dictated not by outputs or metrics that one can measure but rather the vibes passed between managers and executives that have worked their entire careers to escape the world of work.

Until you’re an adult and able to recognise the many ways in which people act deviantly to advance their own interests, you should not be online.

This is interesting to think about and often compelling to use but leaves unresolved one of the first questions posed by chatbots-as-search: Where will they get all the data they need to continue to work well?

David Silver, the principal architect of AlphaGo and AlphaZero. He highlighted a fundamental distinction that seemed to illuminate the entire recent evolution of AI:

“The more human intervention you introduce into a system, the more its performance tends to degrade. Conversely, when humans step back, the system becomes more effective.”

Participants commit to make themselves available for at least 5 hours per week for course readings and discussions.

The course is fully online and open to participants around the world. You will need a reliable internet connection and webcam to join video calls.

The course is free of charge.

Luis von Ahn, co-founder and CEO, took to LinkedIn on Thursday to walk back a previous stance pushing AI use over human employees.

The overwhelming majority (94%) of EdTech Leaders see AI’s potential for positive impact in education, with productivity the highest rated area.  Generative AI (Gen AI) was ranked the top tech priority, with the vast majority (80%) of respondents working in districts with Gen AI initiatives.

When I use AI to build software I feel like I can create almost anything I can imagine very quickly. AI feels like a power tool. It's a lot of fun.

Many AI apps don't feel like that. Their AI features feel tacked-on and useless, even counter-productive.

I am beginning to suspect that these apps are the “horseless carriages” of the AI era. They’re bad because they mimic old ways of building software that unnecessarily constrain the AI models they're built with.

The future of security will move past even the passwordless mixture of factors by adding context, behavior and personal control to dynamically protect identities.

I’ve tried all (46 😵‍💫) AI Coding Agents & IDEs

[Factory, Cursor, Heyboss, Windsurf, Emergent, Wrapifai, Copilot, Lovable, Bolt, v0, Replit, MarsX, Canva, Devin, Github Spark, IDX, Stitch & more]

The most complete list ever made (with demos & notes)