Sources for Building, automation, and the missing layer in AI judgment

Post

Strongest recommendation

Author: Paul Graham.
Recommended by: Jessica Livingston, who argues that universities may struggle to give students freedom to work on their own projects—but that this is exactly what they need to do.
Key takeaway: Graham’s answer is not an entrepreneurship curriculum. Universities should teach powerful ideas and cultivate the ability and habit of building; the two institutional changes he identifies are making startups feel like a viable option and encouraging students to pursue independent projects.
Why it matters: This is an unusually operational founder-formation playbook. Graham ties independent projects to deep learning, finding cofounders, and discovering ideas, then contrasts that with business-plan competitions, which train students to optimize for investor stories rather than user-valued prototypes. His implementation advice is deliberately counterintuitive: return time to students, leave projects genuinely student-owned, and resist adding entrepreneurship deans or innovation centers.

Creator/source: Colossus; the page identifies Dom Cooke as its managing editor.
Recommended by: Patrick O’Shaughnessy, who called Peterffy’s story “fascinating” and linked the profile after a post highlighting him.
Key takeaway: The profile follows Peterffy turning friction into systems: he reduced routine engineering calculations from 20 minutes to 30 seconds, replaced trader intuition with mathematical pricing, and—after a $75,000 loss—rebuilt around fair-value calculations and hedging. He ultimately built what the article calls Wall Street’s first fully automated trading system, despite repeated resistance from incumbent exchanges.
Why it matters: This is a useful case study because automation and risk discipline are inseparable in the story: the same willingness to redesign a broken workflow is paired with explicit controls learned from failure. The result is a more useful founder lesson than a generic success narrative.

Author: Marvin Minsky. Link: No direct book link was supplied in the post.
Recommended by: Amjad Masad, who surfaced the book while responding to a discussion of how impaired emotional processing can leave someone able to reason but unable to make basic decisions.
Key takeaway: Masad summarizes Minsky’s position that emotions are part of human intelligence, not an incidental side effect, and highlights a “selector” for different thinking strategies.
Why it matters: It is a targeted conceptual counterpoint to treating intelligence as the production of plausible options alone. The surrounding discussion makes the unresolved problem concrete: an AI system may produce 20 reasonable answers, while someone—or something—still has to determine which outcome matters.

Building, automation, and the missing layer in AI judgment