What I’ve been reading this week ending 28 September 2025

To vibe or not to vibe

1. Probability: How likely is AI to get things wrong?
2. Impact: If AI gets it wrong and you don’t notice, what are the consequences?
3. Detectability: Will you notice when AI gets it wrong?

AI-Generated “Workslop” Is Destroying Productivity

The science of defiance: A psychology researcher explains why people comply — and how to resist

Alien Oracles: Military Decision-Making with Unexplainable AI

“Calibration by consensus (an example of ensemble learning) uses multiple independent AI agents — perhaps with different algorithms or training data — to analyze the same problem.”

“Calibration by disagreement mirrors the artillery’s adjustment of fire process, where initial shots are expected to miss and their divergence from the target provides essential information for correction.”

“The most impactful AI strategies will frequently defy human logic. The key to cultivating justified trust in these opaque oracles is rigorous calibration and confidence built on experience, not explainability.”

Swarms are For Agents, Not Just Drones

“The difference between assistants and agents is the difference between theater and effects. Assistants help humans type faster. Agents complete missions. They execute bounded tasks, report results, and move to the next objective. They transform tempo.”

Addressing Gen AI’s Quality-Control Problem

1. Conduct an audit
2. Deploy guardrails (simple rules, statistical profiles, AI checking AI)
3. Test the effectiveness
4. Create a learning system (customer model, multivariate experiments, concept tests)

Testing is better than DSA | Ned Batchelder

[DSA = Data Structures and Algorithms]

“…if it will help you get a job, deep-study DSA. But don’t be disappointed when you don’t use it on the job.

If you want to prepare yourself for a career, and also stand out in job interviews, learn how to write tests…”

“Why probability probably doesn’t exist (but it is useful to act like it does)” | Statistical Modeling, Causal Inference, and Social Science

“I agree that “probability doesn’t exist,” in the same sense that “real numbers don’t exist” and “straight lines don’t exist” and “angles don’t exist.” First off, none of these concepts exist in isolation; they’re defined as part of a larger structure. Beyond that, “probability,” “real numbers,” “straight lines,” and “angles” are fictitious; they’re part of mathematical models. Mathematical models are constructs that we use to understand the world. If you draw a “triangle” on the ground with three “straight lines,” then their “angles” will add up to “180 degrees”–except they won’t, because the Earth is not “perfectly flat.” Mathematical models are not reality; they’re tools that we use to describe reality.”

American students are getting dumber — by Matthew Yglesias

“My guess would be that even if schools drop the ball, the best students wind up doing okay thanks to a blend of natural ability, self-motivation, and parental supplementation. But when you hold schools accountable for results at the bottom, they have no choice but to pay attention to instruction methods that work, which has positive results for basically all students.”

Also applies to non-school organizations.

What Makes System Calls Expensive: A Linux Internals Deep Dive

“The point is simple: the cost of a system call goes beyond the small number of instructions that execute in the kernel. It disrupts the CPU’s rhythm by draining pipelines, resetting predictors, and forcing everything to start fresh.”