Horrific but incredible tool to convert a PDF to a podcast episode. I don’t have a key to try it myself, but the examples are pretty incredible.
I guess for people that are ok withh summaries and like getting information via audio, this could actually be useful.
Via Simon Willison
I wish people took that as the lesson rather than “remote work doesn’t work”. I also feel like orgs that suddenly became remote/distributed did it poorly in most cases so have come to poor conclusions.
Interesting comments on language, too. Canonical seems to do this well, even with ~100 countries. I’m not sure why that is.
Somewhat a side note, but I agree this with:
While many orgs have the utopian goal of committing the entirety of institutional memory into neat documentation, knowledge maps, and wikis, I have come to realise that it is essentially a pipe dream. If we look around, rarely is an org able to copy another org’s culture or way of working
The list of good traits to look for when hiring for remote/distributed is good, too.
An interesting theory that great engineering managers are great at debugging. I think my experiences match this, but I don’t think I have enough data (either way) to be sure I believe it or whether it’s more correlation than causation.
Interesting - I’ve been reading a bit about rewilding lately (on The Spinoff and in the linked physical book) and here it is applied to the internet.
I really like that by current job pushes towards this:
Our job is to keep open as much opportunity as we can, trusting that those who come later will use it.
But I think this is one of the biggest challenges:
Instead of tech firms extracting and selling people’s personal data, different payment models will fund the infrastructure we need.
This is a lot of words on “no wrong door” policy, but basically it’s just asking another group for help on behalf of someone (copying them in) rather than just directing them to the other group.
I definitely agree with this approach and try to follow it myself.
Pretty compelling argument to use “login” and “permissions” instead of “authN” and “authZ”.
Good advice on effective communication. I like how the focus is on outcomes rather than specific implementations. Most of the outcomes are ones that I’ve had when communication was working best, too.
This is near the end, but super valuable advice:
One of the most important tools you can use, regardless of channel, is to just decline to engage. If you don’t have anything to add, don’t have time to take on more, or just aren’t interested: say so.
Interesting data on companies with good vs bad blog posts.
Good advice about programming guidelines - this sort of understanding in is what distinguishes junior and senior devs, in my experience.
Scathing look at NASA’s moon landing programme. Good all round, but I love this: “Visionaries at NASA identified a futuristic new energy source (space billionaire egos) and found a way to tap it on a fixed-cost basis”.
Very detailed, good, description of how to write good commit messages - in the GitHub style world, I would say that with docs like this “commit message” is really “PR title and description” - ie. what will (generally) become the commit when the PR is merged. When squashing, the actual commits don’t deserve the same care.
I think one of the reasons I find people calling Gen AI “AI” so annoying is the cycle that Glyoh describes - most of which I remember from being a student doing AI work, then loosely AI dev work.
This is a clever way of circumventing DKIM but using l= has always been a bad idea, and short expiry a good idea, and over signing critical headers (like Content-Type) essential.
Fun story debugging “the wifi only works is when it’s raining”.
Public sector open-source maintenance makes a lot of theoretical sense, but I doubt it could realistically be funded as suggested, and it would subject to awful whims of successive governments. Also, is the public sector capable of remote work?
Good advice on communicating decisions you disagree with to your team - I’ve had to do this a lot, and did not always do it well.
It would be so great if this became a thing: “Specifically, every employer of software engineers should immediately institute the following benefits program: each software engineer should have a monthly discretionary budget of $50 to distribute to whatever open source dependency developers they want, in whatever way they see fit.” Glyph
Interesting thoughts on providing instruction to people using tools - in the poll on power drills, I guessed the right answer but hadn’t thought about it before (despite using one somewhat regularly for years), and probably should have learnt that.
A scathing look at the people behind Google search and its decline. It is fascinating that someone could be the current head of search at Google when they were head of search at Yahoo 2005-2012.
Tempting idea to have LLM generate draft commit messages.
Although this closing point makes me wary:
I have had it hallucinate hilarious things. Never making up changes (thus far), but doing weird shit like adding “Fixed issue #54” at the end.
An interesting write-up on a supply chain attack but I don’t get why all the example requirements files have .tar.gz URLs in them. Who does that for PyPI packages?
This is pretty neat: a SQL schema generated via an SQL query (via Simon Willison)
I like this particular take on the Apple DoJ case but especially the closing:
Microsoft didn’t fall because of the DoJ. Microsoft fell because the web disintermediated Windows. And in a way, Microsoft actually helped that to happen while trying not be disrupted.
And wait. What the fuck are we talking about? Microsoft didn’t fall. Microsoft is currently worth $3.2 trillion dollars.
“There’s a popular notion that young people don’t have the attention spans for in-depth storytelling. That’s ridiculous. Yes, they’ve become hooked on snackable content from Instagram and TikTok. But they’re also quite happy to binge 10-hour factual series on Netflix, or listen to 90-minute podcasts." - So true, and hardly ever acknowledged!
This post on K8s ends positively, but I can’t see how using K8s directly is a good idea unless you’re big enough to dedicate people to it.