After a lot more work than I expected, my video about Recursion is done. Hope you enjoy the journey down the rabbit hole ꩜ youtu.be/YuaJ8x_NcLw

I have a game side project that uses algebra, so I built a small computer algebra system that can manipulate equations. Would anyone be interested in a video walkthrough of how the code works for something like this? I think it's quite interesting 😁

It's kind of funny how a megabyte is either tiny or massive depending on who you ask

A frustrating side effect of search becoming "smarter" in general is that we seem to have lost the capability to just match on a substring. Across the various search boxes I use daily (apple spotlight, google search, github search) they all now fail to find "bar" in "foobar"

I've recently had to write a significant amount of code in the tools that I use for making videos - with no auto complete, no syntax highlighting, no error squiggles, nothing. And I've noticed that the next time I write actual code, I'm way more likely to 1 shot it correctly.

Odd to think that there is now a “vintage” of college degrees that were earned before ChatGPT existed. I honestly wouldn’t be surprised if most companies dropped degrees from hiring requirements before the end of the decade if nothing drastic changes.

Spend enough years working on an artistic or scientific pursuit and you start to realise that the bell curve meme is actually the sine wave meme

When I first released Clay the README was around 2x the line count of the actual library itself. I personally believe that open source, for all it's amazing benefits, has also created enormous pollution in the form of wasting people's time. The issue is complex and has two sides.

I know it's easy to hate on the web as a software platform but every so often I have to fill out one of those turing complete PDFs for some government service in acrobat and it reminds me that there was more than one circle in Dante's inferno.

Nothing quite like the adrenaline rush of undoing 100+ steps and veeeery carefully copying some old code then redoing 100+ steps into the future again. We've all been traumatised by that one time fat fingering ctrl+x and having to redo everything 🥲

There is an insidious idea that has crept into software through the open source movement, which is that software no longer being modified is "abandoned". It's very possible - even desirable - for software to simply be "finished". To serve its purpose reliably, as a tool should.

I find it interesting to think of the work a compiler does as similar to the concept of "simplifying" an algebraic expression - searching for redundancy, combining like terms, replacing identities, etc. Anyone have any academic terms / links for me to have a dig in this space?

The problem with programming language debates is that the critics present the worst possible thing a language is capable of, the supporters present the best possible thing, and precisely 0% of real world projects using that language are either of those things.

Anyone know the name of this phenomenon in human endeavor? Gen 1: Spends career mastering difficult process Gen 2: Learns with difficult process, invents tools to greatly increase productivity Gen 3: Learns with new tools, similar or reduced productivity vs Gen 1

There is so much "mitochondria are the powerhouse of the cell" type trivia in programming - one is that "the stack grows downwards" is always mentioned immediately in stack vs heap discussions. Functionally useless in practise, but a very common question for comp sci exams!

Does anyone know of a "sandboxed real mode" style system, i.e. machine is still running in protected mode, but within each process you have free reign over the entire address space, and peripherals and IO are mapped into special fixed memory regions rather than being syscalls?

You know how explorers love to put their names on things like mountains and rivers? Turns out academics do the same thing, but for ideas. Problem is, inevitably 10 people think of the same idea independently, and then students get taught 10 words that mean the same thing.

Programming jargon can seem intimidating until you figure out that a fancy term like "immutable variable" means "unchanging changeable", and it becomes clear that technical language is just a crowd sourced API design with serious backward compatibility issues

I suspect "what do you think about global variables?" would be quite an enlightening interview question.

I totally agree with the current sentiment of “use a profiler before you try to optimise” - but it’s also important to remember that a profiler won’t reveal quadratic loops, resource contention problems etc that only appear under certain circumstances. Intuition matters too!

Might sound counterintuitive but I think the best advice for improving your high performance programming is to get a much worse computer. It's so easy to commit crimes on an apple M series that you could never get away with on the type of machine most people in the world are on.

It's amusing to me to think that other disciplines like mathematics must have the same type of tribalism as software - there must be an equivalent of the C people, the Rust people, the Haskell people etc in math research

Anyone know of any data / research that explains why the "lose/loose" misspelling became so common online? I find it fascinating - there are so many other more common words that are easier to misspell. Why loose? Possibly driven by some phone keyboard auto correct failure?

The way that people point at large language models like ChatGPT and say “AI is a trillion dollar technology” reminds me of how people used to point at Beat Saber and say “VR is a trillion dollar technology”

It appears to me that all the programming focused AI tools are heading in the direction of "write all the code for me", but does anyone know of any models focused instead on "I want to ask conceptual questions about programming", that can respond with diagrams, references, etc?

I'm working on a video about recursion and I _almost_ caught myself using the fibonacci sequence as a basic example. It's what I was first taught in university, but in hindsight I think it's such a poor starting point for teaching intuition around recursion / recursive functions.

There really is some kind of beautiful, intrinsic relationship between trees and stacks. Sometimes the resonance between two concepts is so strong, I have to assume they're two facets of the same core idea that I can't see just yet...

Nice to still find some humour in programming here or there. Haven't done a huge amount of windows platform dev before, and just stumbled onto the `WIN32_LEAN_AND_MEAN` define 😁

This is fantastic news. When I worked at Canva we had likely one of the largest and most complex typescript codebases, and the compile times were a major productivity hindrance. Best of luck to the team, this will have an enormous downstream impact! devblogs.microsoft.com/typescript/t...

Been doing some lock free programming experiments, and having a lot of fun - Inspired to make a video about atomics. There are lots of little tricks in there, and it’s not a super well known area.

Classic C stockholm syndrome, but it just occurred to me that calling typeless pointers `void *` instead of something like `any *` is quite a poor design decision, and makes the language needlessly confusing. `void` and `void *` are completely different things.

Any naturally high latency computing process (network requests, disk access, waiting for user input, etc) tends to cast a performance "shadow" that extremely wasteful code can hide inside. If you're looking for some low hanging performance opportunities, try looking in the shade!

Be careful of things in programming that are described as "simple". Unfortunately we tend to use that one word to describe two very different things: - Simple (good) achieved by exposing & reducing essential complexity - Simple (bad) artificially achieved by hiding complexity

Once I got down to the nuts and bolts of various platform / OS APIs, I started to realise that just writing the windows / osx / linux platform specific code by hand is often easier, faster _and_ less error prone than trying to wrangle some magical (and leaky) cross platform lib

Laughed when I saw this man page for the `getdents` syscall on linux:

People generally assume attempts at performance optimisation go one of two ways: success, or failure. But often there's a messy third option - the performance did improve somewhat, but now the code is much more complex / brittle. Balancing that trade-off is very difficult 😅

I know it's a cooked idea, but I half wish I could operator overload ==, >, < etc for floats to automatically include an epsilon...

If you've ever wanted to implement your own flexbox UI layout, or if you're just curious how it works under the hood, I've just uploaded a video walkthrough of how Clay's UI layout algorithm works! youtu.be/by9lQvpvMIc

I suppose at some point in the future, someone will inevitably use AI to produce programming videos that copy my style, voice, and potentially even my face as well. It's not going to stop me making the things I make, but it's strange deja vu realising that right now is "before"

You ever read one of those sentences about a piece of software that immediately strikes fear into your heart?

I suspect the way that leetcode et al. let you submit multiple solutions with increasing difficult of test cases actually damages your ability to think through the potential edge cases of a problem on your own. Why try to predict the cases if you know it will just show you?

I know it's very common (and good) advice to say "make your game fun first, before you make it look good", but please don't underestimate how much a little polish early on can boost your motivation keep working on it. Sometimes you need just a little juice to appreciate it.

IME, when a programmer says "I have a problem", there's a high chance that they are describing an issue downstream of the actual problem. There are so many "clever" solutions to something that could have been entirely avoided with a higher level change (of course I'm guilty of this too)

IME, when a programmer says "I have a problem", there's a high chance that they are describing an issue downstream of the actual problem. There are so many "clever" solutions to something that could have been entirely avoided with a higher level change (of course I'm guilty too)

I wish modern programming languages had a clearer way of distinguishing between string literals and runtime created strings. Even with "const" or "immutable" or similar there is always a mechanism to create those at runtime. I want a reliable "isStringLiteral(string)" function.

A comment I often see is "But doesn't the compiler just do that optimisation for you?". You'd be surprised at how little wiggle room the compiler has, or how tiny changes to your code can drastically affect the output. Godbolt + basic ability to read asm will change your life🙂

I wonder how much total compute power has been spent over history skipping over the "." and ".." entries reported in every directory listing by the posix and windows APIs...

does anyone else's computer desk look like this at basically all times

It's funny how the word "edge case" has evolved from "the user submitted 10 million print jobs in one second" to basically mean anything that isn't the absolute happy path. "The edge case where the user needs to reset their password" brother please I'm begging you

In the interests of transparency and for other creators out there, I just spent 30-40 hours working on a video that didn't pan out. I just couldn't get the logical throughline to feel smooth. Have started it again from scratch and I think I've got it, it happens sometimes!