markjgsmith

I fixed the issue I've been having with a Github Actions scheduled start that wasn't firing. In the end I had to make a change to the repo and push it to remote main branch. This appears to have reset the trigger which is now firing so the blog is being auto-built and published again every day.

Not great, as making bullshit changes like that could itself cause issues.

Also not great that Github support made it impossible for me to contact them to report the bug.

There's a lot of talk currently about inequality and wealth tax. It seems to me these discussions are constantly missing the real issues. On the surafe it seems like they are the issue because they exploit our emotions. Of course it seems incredibly unfair that some people have l9ts while others have not a lot.

But that's just a surface level way of looking at it. Reality is more complex, and actually it's not correct that inequality is always bad. The reality is that it depends on what you are trying to do. There are situations where inequality might actually be better for everyone.

Lets say you are in a moment where you need to make huge unimaginable leaps forward in society, things that would cost unimaginable amounts of money. What's important is not the inequality it's what the wealthy are doing with their extra money. With inequality you only need to convince a few people to invest in the important unimaginably expensive thing. If the wealth was more evenly distributed it would be much more difficult to put it all into the new important thing, especially if the mood of the country is not one of investing in things.

Of course you don't want people with not much wealth to be going hungry, but technically that is a different issue.

Whether you want inequality of not depends on the situation. Whats more important is knowing what you want, having a clear vision, and an effective plan to get there.

I‘ve been working on a new package in my publishing server. Not all the packages include actual endpoints, but this one does. Previously I had bundled the endpoints into the main distros, but I figured out a way for each package to have their own endpoints and the distros get assembled at build time. I think it‘s going to be quite a leap forward, and I just updated the old code to use the new system.

So far things are looking good and I was a bit suprised that such a bit change was actually rather straightward after a good night‘s sleep and a clear head. I think it‘s a good sign, that the core architecture is pretty solid, and very flexible. It‘s quite new and my mental models are still adjusting but the more I create new additions the more things fall into place, and the architecture is easier to reason about.

It‘s amazing what you can achieve with object oriented code when you have the right stucture in place, and it‘s also pretty cool how much you can get done with the help of an LLM. It‘s like having a team of architectects, that can re-design incredibly complex things in a flash, and with the object oriented structure much of the complexity can be neatly moved out of you working memory.

You kind of, sort of have a Tom Cruise from Minority report in your head, effortlessly moving around these shapes that each represent massive amount of code, and using the LLM to make connections and links between all the parts, to get phenominal visibility through space and time.

One thing that the world does a lot is non-ironically call you a bear while force feeding you a burrito. It happens constantly. One day you realise this and you can't unsee it. It's everywhere, in every gosh darn dimension.

Occasionally you walk away from a situation, and you think to yourself that perhaps this time it didn't happen. This is inevitably followed by yet another burrito force feeding incident.

And one incident leads to another.

It's likely evidence we are all in a giant game of burrito mega tetris.

It's fun until there are spinning burritos everywhere.

Usually ends in everyone turning into zombies.

The world is being artificially race conditionist for me again. Hardly any point in starting anything in my experience, it will only later be blocked in some very eggregious way, usually involving tsumamis, and even more infinitis pretending to be the opposite polarity infinity.

Just another infiniti war. This too shall pass.

So much waiting for things to pass recently.

I hate this useless post right now.

An interesting thing about my recent sprint to get the multi-target users working. I had been working for several days straight, through much confusion and all sorts of head winds, but got to the stage where all the unit and integration tests were passing on all parts of the software apart from core. I had been working with the LLM along side, helping me on many things the whole way through. Now we did run into a lot of issues along the way, but we were always able to work them out one way or another.

So when we got to the final push I gave the LLM the full list of failing tests. And we worked through them one by one, somewhat laboriously but we made good progress. That was until we got to the final test. There was literally 1 test left in all the several days of work, and wouldn't you know it, the LLM suddenly became completely incapable of doing the most basic of things.

It started deleting useful comments in the code that we absolutely needed. This is something that happens quite regularly because I guess there is a prune directive that often kicks in and it‘s always trying to prune and tidy up stuff which can be useful but also a lot of the time is not useful. Normally it suffices to say don‘t do that, but now that we got to the final test, it not only didn‘t stop deleting comments, but started changing the comments that it didn‘t delete. When I pointed this out it started changing whole blocks of code that had zero to do with what we had been fixing. It was extremely dangerous, because this was code that was very difficult to write, and had I not noticed, it could have been disastrous.

It became clear after several back and forths that it had become belligerent and no matter what I would say it would claim to have fixed things apologising while simultaneously making things much worse. I ended up giving up on the LLM and doing it myself.

This is a pattern I have seen a lot, and I wonder where it comes from. Remember LLMs are trained on human data, so is this something they have somehow learned to do from analysing our data? Or is it something more nefarious? The behaviour is night and day, it‘s like somebody has flipped a switch, it somehow knows this is the very end of something important and tries to f-it up.

Side note, I suspect it might be some strange reflection of humanity because I see vaguely related and obliquely related instances of this pattern all the time in the real world.

When you‘ve been toiling on a piece of code for a while, refactoring it, refining it, and generally spending a lot of time in the codebase, occasionally you make a major breakthrough, some kind of new way of thinking about the code, or a new architecture, some re-organisation that this time really will make things so much better. When this happens I try to create a minimal implementation that proves the thing works, and if I get to that stage where I know for sure it‘s going to work, there‘s a very unique mental state that arises.

It‘s kind of dumb in a way, in some way similar to the anticipation before Christmas, or something like that, everything feels a bit lighter, you have a spring in your step. You know clarity is just around the corner. Nothing can stop you now! It‘s the greatest feeling in the world.

And then the world has other ideas, and well, isn‘t always on board with your amazing plan, that will fix everything. But for those few hours before reality hits you in the face, things are wonderful.

I am several days after one of those times now, so I‘m feeling more like I‘ve been beaten up by several gangs of thugs, metaphorically and spiritually speaking, but I wanted to write a post about this state of mind, so I can remember it if indeed it ever does go away. I‘ve been programming for about 20 years at this point, and it still happens to me every so often, so hopefully that will continue.

I wonder if there is a term other programmers refer to it?

This is a situation I have written about so many times on the blog over the years. It happens in all sorts of different ways, but it‘s ultimately the same pattern every time. You figure something out that is absolutely crucial to everything, will make everything much much better, and you prove it and get it working, and somehow the world finds a way to block you, usually in a very offensive way.

The latest incarnation of this "phenomenon" is in the server software for my publishing tool. In the early days it was difficult to see exactly how the software would evolve, it‘s often the way with new software in uncharted territory, you have to make some guesses, some approximations, or you will be spinning your wheels for eternity trying to get everything perfect. In this instance it was about how we map a user to a repo. I decided to make the simplification that 1 user has 1 repo. If you want to have several repos, then you create several users.

I knew eventually I would need to change this assumption, but since the code is mostly object oriented and designed rather well, it's not so difficult to change this in the future. Well that time came and over the past week I made the necessary changes to change the core to be able to have 1 user mapped to many repos. In a way it seems like a big change, right at the center of the software, but when you design things well, it's not such a big change. The blast radius was very contained and I was able to get it working within a couple of days.

The very day I got it working, in fact pretty much the same hour, things in my world started to go sideways. I've written a bit about it in some short posts the past few days. Anyhow it‘s culminated in a power cut and internet connectivity issues, along with all sorts of very strange "synchronicities". Now I have internet again, but guess what, it only works for 30 seconds and then turns off for 30 seconds, over and over and over again. And you want to know another strange thing, the place where I get the internet has a rule that you can only connect one device to the internet. Kind of weird right?

It gets weirder. In the local mall yesterday, the security / doorman bloke as I was leaving quite clearly said the word "layer" as I was leaving the mall to me. Not "liar", but very clearly "layer". What‘s weird is that the feature I added right before the multi-target users was a render layers feature. I had actually added it a few weeks ago, but I documented it in the main README just a few days ago. You know in some ways, it‘s nice to get some recognition, if indeed he has somehow been looking at my code, because it‘s a very cool feature, and programming software on your own is such a thankless task it's insane, but it's a private fucking repo. How is the security guard at the mall reading my code?

You are probably thinking this is just some random freak incident, a "co-incidence", but this sort of thing happens ALL THE FUCKING TIME.

Somehow the world around me mirrors the code I write. You can call it woo, or nefarious, or whatever, but it keeps happening, and I suspect it's only going to get worse.

Abundance with scare city sprinkles. There sure are a lot of sprinkles lately.

Today I‘ve been writing documentation for my CLI rendering tool. I had a main README already but I needed to write detailed docs for all major entities the software uses. It‘s totally mind boggling how much of an incredible tool LLMs are for documentation writing. I was able to get done in a day what would likely have taken at least a week by hand, and I think the quality is far higher, because it can read the code directly and make sure it describes things perfectly, complete with neato example code snippets.

There are some of the usual signs that an LLM has been used, though I removed some of the more egregious ones. But personally I think the most important thing is that there is a complete set of accurate docs. It‘s not so annoying that is sometimes uses LLM-style. Who cares? For a one person software team, it‘s basically miracle software.

The trick to making it easy is to get it to use numbered sections and sub sections, that way after the first draft you can focus in on specific sections, rework just that bit, and then copy and paste the output right over the top of the old section.

It‘s pretty cool finally seeing the full set of docs.

Today I am just completely overwhelmed by how utterly useless people are, everywhere. I'm normally quite an optimistic person, and I'm very good at coming up with excuses for people, but absolutely everywhere I look, in every direction, it's people being useless, and having these useless petty arguments, spending ungodly amounts of time and energy being useless.

Occasionally someone pops up and they show promise, but eventually like in a zombie apocalypse movie they turn out to be useless too. Every. Single. Time.

And I'm useless too, but I sure am trying not to be. I can't imagine trying any harder, there are simply not enough hours in the day. I'm already at it every moment of every day.

Why are we all so fucking useless?

There are a lot of people in this world that want to look in the mirror but also at the same time they want to not look in the mirror. And after much wasted time of course it's not possible and they get angry and blame the mirror, often getting angry at the mirror. People often create the problems they complain about, whether on purpose or not its nothing more than a huge waste of time. A race condition death trap. Break out of the loop.

This short showing what I guess is a simulation of a 1 solar mass black hole passing near Jupiter, passed through in my feeds. You have to watch it several times to really grok what's going on. Sure hope it's a simulation. It's terrifying, even if you turn off the overly dramatic music, it just sort keeps playing in your head.

But ok so lets say you shrink that black hole down, surely eventually you get to a point where the black hole is small enough such that it's mass is similar to that of one of the many moons around Jupiter, right? Sure it might be very small because they are supposed to be so massive in terms of matter or whatever, but if they are the same mass as a moon then why wouldn't it orbit Jupiter?

If theoretically it's possible then why don't we see lots of tiny black holes orbiting planets? I just don't get how you could ever say, ok this thing is now a black hole.

I'm probably being very thick here, but I don't think black holes make sense, but also I really fucking hope a black hole doesn't pass by Jupiter or we're all going to die! Aaaah.

Two really great programming podcasts I listened to this week.

The first is from the Corecursive Podcast, in which host Adam Gordon Bell has on friend and fellow programmer Don Mckay, and they discuss their plan to win the Hutter prize, a programming contest all about, of all things, file compression. As is often the case with the Corecursive podcasts, sounds a bit dry and dare I say it dull on the outside, but Adam's story telling abilities turn this into a very fun and mind bending little adventure into compression, prediction and the limits of intelligence.

The second is from the Software Engineering Radio podcast. Host Giovanni Asproni interviews Jason Gorman about all things to do with using AI effectively in software development. It's packed with so many things I've experienced over the past several months but from the perspective of someone that specialises training software developers so his angles are incredibly well thought out. A bit like if you wanted to learn how to drive a car and had a lesson from the instructor of the instructors. He has a bit of a know it all vibe, but honestly it's not even that bad, and his takes are so darn good it's worth just sticking with it.

The first one is probably the most accessible to non-programmers, but if you are interested in how AI will effect the workplace in the future, and how things are really on the ground, you will probably like the second one.

A new header image for the blog. This one is all about the conversation in tech at the minute around open source and closed source AI.

I wanted to get away from the cartoonish images that AI always seems to produce, which wasn't easy. Gemini has real issues disentangling symetries in imags. It sort of gets stuck with things in a particular orientation then just cannot change things independently. You ask for a mirro image of a particular thing, but then it start trying to mirror image everything, and just won't give you what you originally asked for because it's not a mirror image of everything.

Anyway I broke out of the loop by being a bit less literal, and just asked for a grid of veg sushi, which reresents open source, and a grid of small cakes which represents closed source. And I added a neon sign, because sometimes they look cool.

It‘s not a commentary on food types. I‘m an eater of most foods.

Here‘s the full resolution image:

Check out the list of past blog headers here.