Processor boxes list cores and threads, and programming courses warn about them early. What is multithreading (also written multi-threading)? It is one program doing several things at once by splitting its work into…
A place to write, and be read.
Practical guides from vlvd.net on AI safety, security testing, study skills, history and games. Read them, follow the writers you like, or write your own.
Latest posts
Fresh off the bench.
Newest first, from everyone. Sign in to follow people and see only their posts.
"Open it in a sandbox" is common advice for a file you do not trust. So what is the definition of a sandbox in cybersecurity? A sandbox in cybersecurity is a closed-off space where a program can run without being able to…
Every computer guide says to keep your firewall turned on. But what is a firewall in a computer network, and what is it actually doing? A computer firewall is a gatekeeper for network traffic. It sits between your…
Computing lessons and spec sheets talk about buses, 32-bit and 64-bit systems, and data moving "over the bus" as if everyone knew what that meant. So what is a data bus? A bus is the set of shared lines that carries data…
News stories say a botnet knocked a website offline, or that thousands of home cameras were hijacked at once. So what is a botnet? It is a network of computers and other devices that criminals have taken over without the…
You see the word in update notes, in error messages and in talk about phones and servers: the kernel. So what does a kernel do? In short, it is the part of the operating system that runs the computer itself. It decides…
Every program you run spends much of its time waiting for data. A CPU cache, also called a processor cache or CPU memory cache, is a small, fast memory next to the processor core that keeps copies of data from frequently…
Most students who take a practice test look at the score, feel good or bad about it, and move on. That throws away half of the value. Feedback on practice tests, meaning finding out the right answer and why, straight…
If English is not your first language, an English textbook asks you to do two hard things at once: understand the ideas and decode the language they are written in. When a practice question goes wrong, it is hard to tell…
Most students know they should not cram, but very few know what to do instead. Spaced repetition for exams means you review the same material several times, with gaps between the sessions, instead of piling all the…
The usual way to study is to read first and test yourself after. Pretesting flips that order: you try to answer questions on a topic before you have learned it, get most of them wrong, and then study. It sounds like a…
Open almost any textbook and the practice problems come in blocks: twenty questions on one method, then twenty on the next. It feels efficient, and it is how most of us were taught. Interleaving practice does the…
Multiple-choice exams look easier than essay exams, because the right answer is printed on the page. That is exactly why they catch people out. The wrong options are written to look right, and if you only half know the…
If you have ever stared at a labelled diagram until every arrow looked familiar, then blanked on it in the exam, you already know the problem. This guide shows how to study diagrams by turning them into quiz questions…
A study group can be the best hour of your week or the least useful one. Too often it turns into reading the notes out loud, or chatting with the books open. The fix is simple: quiz each other. When every person has to…
The night before a test, many parents do the same two things: read the notes aloud, or ask "do you know it?" and accept "yes". Neither helps much. If you want to help your child study for a test, the most useful role you…
"Ask yourself why" is one of the most common pieces of study advice. It even has a technical name: elaborative interrogation. The idea is that when you read a fact, you ask why it is true and answer in your own words,…
You reach the end of a textbook chapter, the pages felt clear, and you are fairly sure you got it. Then a question in class or in the test shows that you did not. The way to check your understanding is not to ask…
Today, much of AI training depends on people judging whether an answer is good. That works while people can tell. Scalable oversight is the research problem of keeping that check working when the AI is better at the task…
If you read about AI safety for long, you meet two terms that sound almost the same: outer alignment and inner alignment. Outer vs inner alignment is not a matter of style. They name two different places where a trained…
A trained neural network does useful things, but nobody wrote down how. Its knowledge sits in millions or billions of numbers. Mechanistic interpretability is the research effort to read those numbers: to explain what a…
Ask what a chess program, a disease-research system and a question-answering assistant have in common, and the obvious answer is very little. Instrumental convergence is the idea that, despite their different goals, all…
Whenever we train an AI system, we give it a number to push up: a score, a reward, a rating. That number is a stand-in for what we actually want. Goodhart's law in AI is the observation that pushing hard on such a…
An AI system can learn its skills perfectly and still learn the wrong goal. That is goal misgeneralization: a trained system keeps its abilities in a new situation but uses them to pursue something other than what it was…
Every machine should have an off switch. For a simple machine that is easy: it has no opinion about being switched off. Corrigibility asks how to keep that property in an AI system that pursues goals, because a system…
Most AI assistants learn their manners from people who rate thousands of answers. Constitutional AI swaps most of that rating for a short written list of principles, and lets an AI model apply the list to its own…
AI safety via debate is a proposed way to supervise AI systems that may know more than we do: two AIs argue opposite answers and a human judges the winner. How do you check an answer from a system that knows more than…
Before an AI model is released, someone has to try to make it misbehave. AI red teaming is that job: deliberately attacking a model with tricky, hostile or unusual inputs to find harmful behavior before real users do.…
Most siege stories are told from outside the walls: the army arrives, the machines go up, the stones fly. But how did castles defend against a siege? The short answer is preparation. A castle that held was usually one…
Trebuchet, mangonel, ballista, battering ram: you meet these names in history lessons and castle tours, often without a clear picture of how they worked. Medieval siege weapons were machines built to do one of four jobs:…
Write here too
Your own page.
Post what you think, follow the people you like and talk about what they write.