And from February, Stephen Wolfram’s tome on ChatGPT. This is definitely not gentle, but also very good.
High-level introduction to how LLMs work.
I’m not sure this is really a ‘gentle primer’, but I do think this is a very good introduction.
An approach to window management unlike anything I’ve seen. I really like the idea of a mosaic as a sort of multi-dimensional tiling.
Via Lukas Mathis
I must admit I’ve never given the tech debt metaphor much thought; it’s always framed as an unmitigated negative, something outside factors foist upon you, and that’s how I’ve tended to think about it.
This is a deep dive into the metaphor, examining how it works on multiple levels, including how you can intentionally taking on tech debt, using the time saved as an investment to ship faster.
Via Simon Willison
Glyph describes some concrete steps to follow to begin splitting a monolith into a micro service (steering well clear of suggestion that micro services solve everyone’s needs).
This is a, to me, novel concept: Defence of design decisions or ‘why did we do it like this’.
I have never been involved with a large-scale project that didn’t have seemingly-weird decisions (and I’ve contributed my fair share), and an explanation would probably have helped.
The only problem I see will be in identifying them before it’s too late (and you forget why), but the presence of such a document surely shouldn’t hinder that.
In a sense this is the “comments should explain why not what“ maxim extracted into a document.
Via Hillel Wayne
(Paywalled at the Athletic.)
Reads like a love letter to the greatest player ever.
Interesting thoughts from Hillel Wayne on how to get more from syntax highlighting.
Being context-aware is usually a good thing; the difficulty often arises when determining context: either it’s done automatically, with all the errors inherent in trying to guess that, or it’s a manual process, at which point it becomes something that you have to get in the habit of using (or not, as often ends up being the case). But I definitely think there’s something to the idea.
Interesting deep dive into popular modern applications’ performance from a perspective of offline-availability, ownership and longevity, with a proposal for a new (to me) model, Conflict-free Replicated Data Types (CRDT).
I remain unconvinced that something like the hyper-collaborative experience of Google Docs can be replicated under anything except an always-online paradigm; on the other hand, I have yet to see that experience used productively – it’s a good demo, but not very useful in my eyes.
I’m late to the game, but Netflix recently added a long overdue toggle to prevent those blaring previews.
This also prompted me to check for YouTube, and they have a similar toggle, although it’s for some reason only available on iOS [and iPad OS, presumably].