physicles

Every now and then someone shares a small web link here (Kagi is one aggregator). It’s like survivors picking up the shards of civilization after the apocalypse. Of course such a project can remain viable and useful as long as it remains niche, which is virtually guaranteed as long as there’s no money in it.

It’s incredible how too much money corrupts everything it touches.

I switched to Linux in 2018, and I hear you on the quirks. But now, an LLM can fix basically any quirk you hit. I’ve been surprised multiple times. I’m also on Pop OS, which feels more batteries-included than Ubuntu.

I’d guess that’s what GP was referring to. I’ll also be a day 1 preorder.

There are differences across vendors, but there’s a way to build with k8s where the benefit far outweighs the cost.

We run a bunch of services in two very different cloud vendors (one of which used to be DIYed with kubeadm), and also on dev machines with k3s. Takes a while to figure this out and to draw the kustomize boundaries in the right place, but once you do, it’s actually really nice.

Two things work in our favor:

- we’ve been at this for around 8 years, so we didn’t have to deal with all the gotchas at once

- we aggressively avoid tech that isn’t universal (so S3 is OK, but SQS or DynamoDB is not; use haproxy instead of ingress controllers; etc)

That part was really surprising to me because for the kind of compute lake he’s talking about building, k8s seems like a pretty good fit for the layer that sits just above it.

We run k8s with several VMs in a couple different cloud providers. I’d love it if I could forget about the VMs entirely.

Is there a simpler thing than k8s that gets you all that? Probably. But if you don’t use k8s, aren’t you doomed to reimplement half of it?

Like these things:

- Service discovery or ingress/routing (“what port was the auth service deployed on again?”)

- Declarative configuration across the board, including for scale-out

- Each service gets its own service account for interacting with external systems

- Blue/green deployments, readiness checks, health checks

- Strong auditing of what was deployed and mutated, when, and by whom

Was just listening to the Lenny’s Podcast interview with Simon Willison, who mentioned another such incantation: red/green TDD. The model knows what this means and it just does it, with a nice bump in code quality apparently.

I’m trying out another, what I call the principle of path independence. It’s the idea that the code should reflect only the current requirements, and not the order in which functionality was added — in other words, if you should decide to rebuild the system again from scratch tomorrow, the code should look broadly similar to its current state. It sort of works even though this isn’t a real thing that’s in its training data.

I have a Supernote and was looking at different models for handwriting recognition, and I agree that gemma4-26B is the best I’ve tried so far (better than a qwen3-vl-8B and GLM-OCR). Besides turning off thinking, does your setup have any special sauce?

This is in fact what happened with SpaceX. Musk had the vision to make first stage reusability a thing with the Falcon 9, and Gwynne Shotwell is the operational genius behind the juggernaut that it is today.

Hilarious quote from her Wikipedia page:

> Shotwell has received particular praise from NASA Administrator Bill Nelson for her "phenomenal" leadership of SpaceX as it developed the Falcon 9 into the "workhorse" of the space launch sector.[19] Nelson had reportedly been concerned in 2022, after Elon Musk purchased Twitter, that it would be a distraction at SpaceX, but became more comfortable after meetings with Shotwell left him feeling reassured that she was in charge of day-to-day operations.

The idea lingered for longer than that. China under Hu Jintao wasn’t exactly friendly to the west, but it was Xi who really set China on its present course to build a multi-polar world, make real noise about reunification with Taiwan, etc.

This new direction didn’t become clear to both sides of the aisle in the US until a year or two into Xi’s tenure. If someone else other than Xi had been chosen, we would likely have a very different China today.

For sure. I feel I need all of my experience to discern the difference between “slightly different, and should be combined” and “slightly different, and you’ll regret it if you combine them.”

One of my favorite things as a software engineer is when you see the third example of a thing, it shows you the problem from a different angle, and you can finally see the perfect abstraction that was hiding there the whole time.

I’m one of those that have thrown out Postel’s law entirely. Maybe the issue is that it never defines “strict”, “liberal”, and “accept”. But at least for public APIs, it never made sense to me.

If I accidentally accept bad input and later want to fix that, I could break long-time API users and cause a lot of human suffering. If my input parsing is too strict, someone who wants more liberal parsing will complain, and I can choose to add it before that interaction becomes load-bearing (or update my docs and convince them they are wrong).

The stark asymmetry says it all.

Of course, old clients that can’t be upgraded have veto power over any changes that could break them. But that’s just backwards compatibility, not Postel’s Law.

Source: I’m on a team that maintains a public API used by thousands of people for nearly 10 years. Small potatoes in internet land but big enough that if you cause your users pain, you feel it.

Vantage looks great! I’ll try it out this weekend.

To do the job that swarf does, I found that the bwrap sandbox I’d been using is the perfect place to mount a folder to catch markdown junk and keep it out of the project’s actual git repo. Works great.

He does go on to say “this is basically impossible if you want to live in society”, but yeah the order in which the information is communicated isn’t optimal for skimming.

Interesting, I didn’t know Dostoevsky also made this popular (unless you’re just tongue-in-cheek ignoring the earlier source, which is the Bible)

Installed Carbonite on my parents’ computer something like 15 years ago, and it still works (every now and then my dad tells me he used it to recover from a bug or a mistake).

But I have no idea where the company currently sits on the spectrum from good actor to fully enshittified.

In Slack it can get even worse.

If you turn on Markdown formatting, shift+enter adds a new line, unless you’re in a multi-line code block started with three backticks, and then enter adds a new line and shift+enter sends the message.

I can see why someone thought this was a good idea, but it’s just not.

Well there went a half hour of my day. What a fantastic read.

This is all valid. Your original comment came across as a troll because it implied that nobody could ever feel good about stuff they built with AI. Asserting that you know more about the emotional state of strangers on the internet than they know themselves is arrogant.

Thank you for sharing this! I love how writing prompts like this forces us to clarify our own values and sense of engineering taste. I’m really curious to see what an agent would find in our code base with this.

Mind sharing that prompt? This is one of my favorite uses for AI too, but I’m just using it to fix the stuff that’s already top of mind for me.

You’re missing the group of high performers who love coding, who just want to bring more stuff in the world than their limited human brains have the energy or time to build.

I love coding. I taught myself from a book (no internet yet) when I was 10, and haven’t stopped for 30 years. Turned down becoming a manager several times. I loved it so much that I went through an existential crisis in February as I had to let go of that part of my identity. I seriously thought about quitting.

But for years, it has been so frustrating that the time it took me to imagine roughly how to build something (10-30 minutes depending on complexity) was always dwarfed by the amount of time it took to grind it out (days or sometimes weeks). That’s no longer true, and that’s incredibly freeing.

So the game now is to learn to use this stuff in a way that I enjoy, while going faster and maintaining quality where it matters. There are some gray beards out there who I trust who say it’s possible, so I’m gonna try.

100% agree. Velocity at level 8 or even 7 is a whole order of magnitude faster than even level 5. Like you said, identifying the core and letting everything else move fast is most of the game. The other part is finding ways to up the level at which you’re building the core, which is a harder problem.

I’ve had a couple wins with AI in the design phase, where it helped me reach a conclusion that would’ve taken days of exploration, if I ever got there. Both were very long conversations explicitly about design with lots of back and forth, like whiteboarding. Both involved SQL in ClickHouse, which I’m ok but not amazing at — for example I often write queries with window functions, but my mental model of GROUP BY is still incomplete.

In one of the cases, I was searching for a way to extract a bunch of code that 5-6 queries had in common. Whatever this thing was, its parameters would have to include an array/tuple of IDs, and a parameter that would alter the table being selected from, neither of which is allowed in a clickhouse parameterized view. I could write a normal view for this, but performance would’ve been atrocious given ClickHouse’s ok-but-not-great query optimizer.

I asked AI for alternatives, and to discuss the pros and cons of each. I brought up specific scenarios and asked it how it thought the code would work. I asked it to bring what it knew about SQL’s relational algebra to find the an elegant solution.

It finally suggested a template (we’re using Go) to include another sql file, where the parameter is a _named relation_. It can be a CTE or a table, but it doesn’t matter as long as it has the right columns. Aside from poor tooling that doesn’t find things like typos, it’s been a huge win, much better than the duplication. And we have lots of tests that run against the real database to catch those typos.

Maybe this kind of thing exists out there already (if it does, tell me!) but I probably wouldn’t have found it.

I’ve wondered for years if this could be quantified. Three orders of magnitude totally justifies the cost, if you care about science.

Astronauts are made of different stuff. Truly the best of the best.

I was just trying to find a blog post that I read years ago where someone wrote about storing their furniture at IKEA. Couldn't find the post, but the idea helped me downsize during a recent move.

Love it. This is so clearly the way to solve the jq writeability problem. I’m going to replace jq with this immediately.

I’d love for it to be true that Trump isn’t just a narcissistic buffoon. Where are you frequently finding evidence of this?

I’m curious, do you agree with the statement, “it would be better for this personal wiki not to exist, than for it to have been built with AI”?

Because without AI it probably wouldn’t exist.

Heh. Trump asks the oracle at Delphi what will happen if he launches the war.

“The war will surely achieve regime change,” replies the oracle.

“Great, let’s go,” says Trump, who never read Herodotus.