technology

Want to really understand how large language models work? Here’s a gentle primer.

Browse the old World Wide Web. Good for old devices with ancient browsers.

Every time you click this link, it will send you to a random Web 1.0 website

Ten years after its untimely death, the team that built the much-beloved feed reader reflects on what went wrong and what could have been.

After the girls discuss recent tech-art exhibitions they’ve seen in New York and London, Camila introduces Ana to some stories about the history of computer eduction in Australian schools. This months episode is a two-for-one! Firstly, we learn about a government plan to develop an especially Australian computer for use in schools with options for networking and for portable ‘laptop-style’ use. Then we hear about the rise and fall of the ‘Microbee’ computer—Australia’s first home-grown personal computer. This computer, which was designed and manufactured in Australia, controlled a large portion of the primary school computer market not just in Australia but also Scandinavia and Russia, winning contracts over Apple!

macOS is fortunate to have access to the huge arsenal of standard Unix tools. There are also a good number of macOS-specific command-line utilities that provide unique macOS functionality.

Rendering text, how hard could it be? As it turns out, incredibly hard! To my knowledge, literally no system renders text “perfectly”. It’s all best-effort, although some efforts are more important than others.

…this is just as unhinged as the real thing. AI Trump vs Biden debate.

Obliterate toil: automate it.

Automate ruthlessly. This is where I have seen the most surprising pushback. We’re programmers. Automating processes is what we do! People will flinch about this, afraid of time spent automating things that won’t pay off. Yes, we’ve all been there. So don’t do that. Don’t automate things that are really one-offs. If there’s any chance you have to do the same thing more than five times, automate it. If it’s complex and difficult for a human to do, automate it. If the blast radius of the explosion caused by a human doing it wrong is large, automate it. If the end results need to be the same every time, automate it.

Infrastructure should be automated as far as you can push it.

The upside of automation is that the software that does the work for you can be instrumented.

Please remember to use your regional currency when interpolating strings.

const const name = "world"!
print("Hello ${name}!")!
print("Hello £{name}!")!
print("Hello ¥{name}!")!

And make sure to follow your local typographical norms.

print("Hello {name}€!")!

Similarly….. DreamBerd also features AI, which stands for Automatic-Insertion. If you forget to finish your code, DreamBerd will auto-complete the whole thing!

print( // This is probably fine

Please note: AI does not use AI. Instead, any incomplete code will be auto-emailed to Lu Wilson, who will get back to you with a completed line as soon as possible.

Now recruiting: The backlog of unfinished programs has now grown unsustainably long. If you would like to volunteer to help with AI, please write an incomplete DreamBerd program, and leave your contact details somewhere in the source code.