Leon Mika

Tue, 18 Aug 2026 07:11:00 +1000

Exit 1.

Mon, 17 Aug 2026 16:38:23 +1000

It’s said that maths is not so much invented as it is discovered. I think the same thing could be said for most software. Sometimes a design is clear to everyone. You just need to “talk through” to the discovery.

Mon, 17 Aug 2026 10:17:17 +1000

There are some things you wouldn’t have thought you’d miss when you loose a contact that worked at a library. Access to old newspapers is one such thing. Now I’m calling branches saying things like “can I collect your old newspapers? I have a need for old newspapers. Can I have them?”

Mon, 17 Aug 2026 07:37:57 +1000

Forget national service. Maybe what we need to introduce is that everyone is required to be a garbo for a month. That way, everyone will think better about throwing their litter on the ground.

Sun, 16 Aug 2026 21:12:15 +1000

Silhouette.

Creating Keyboard Split Tracks In Logic Pro

Sun, 16 Aug 2026 10:04:21 +1000

Here are instructions for how to create a “keyboard split” track in Logic Pro, where the left hand plays a different software instrument to the right. I used to create an electric piano patch with tones above C2 (1 octave below middle C) having a delay.

Create the individual tracks as you would normally.
Select them, right click, and select “Create Track Stack.”
Select “Summing stack.”
In the track settings on the left, adjust the “Key Limit” so that one track covers the left hand and the other covers the right.
Select the track stack to record a take.

I don’t know if this is the correct way to do it, but it was surprisingly difficult to find this information online: many of the queries for “split track” were for splitting clips. But this seems like a pretty versatile technique for combining software instruments and having them respond to a single input source. Based on what’s available, it looks like it’s possible to do something similar across the velocity dimension too.

Sun, 16 Aug 2026 08:09:25 +1000

I’m trying to get rid of my air rowing machine. Anyone interested? Free if you’re willing to come and pick it up. Details in the linked ad.

Update: item has been collected and is no longer available. Good thing too: I didn’t need to dig out the marquee tags for this.

Sat, 15 Aug 2026 21:24:38 +1000

Short(-ish) walk along the Yarra Trail this afternoon, near Heidelberg.

Sat, 15 Aug 2026 08:28:04 +1000

🔗 The Imperfectionist: Let’s stop talking about A.I.

AI’s the core subject of this article but it was this part about overthinking that got me, well, thinking:

Whatever the details, this is where overthinking kicks in, because it feels like the solution to the anxious feelings in your body is to scurry off into your head and think some more, read some more, argue some more, maybe use A.I. some more.

I see this in a few people I know, including myself. And it definitely is a protection mechanism. The quote comparing it to an addiction was a powerful realisation to me.

Via: Sal’s Blog

Sat, 15 Aug 2026 07:57:45 +1000

Congratulations to Marco Arment for the release of ~~The Reminders~~¹ Unforgetful. I’ve enjoyed listening to his dev stories on ATP, and I’m glad he managed to release it. Looks like a nice little reminders app.

That’s a reference, John. 😉 ↩︎

Fri, 14 Aug 2026 22:20:35 +1000

🔗 Noēma: The Future Of Knowledge Workers In An AI-Driven World

A thought-provoking essay. Not sure I believe my role to be utterly pointless at this stage. We know whom we serve, and see it appreciated. Yet there is still a slight malaise I’m feeling that I can’t quite put my finger on.

Fri, 14 Aug 2026 08:53:05 +1000

I’ve been using MonoLisa for about a month and a half now. It’s been so nice to use I’ve went ahead and bought a license. I recommend trying it, if you need to deal with a monospace font in any capacity.

Thu, 13 Aug 2026 17:32:46 +1000

Also, few things in life are so urgent that it couldn’t sit for a little while. A problem that’s been brewing for a number of years can wait a few more frickin’ hours.

Thu, 13 Aug 2026 17:29:56 +1000

Learnt an expensive lesson today: the price of a job should be based on how much those who are willing to do it would take. It should not be based how much you’d ask for if the job was offered to you. Ah well.

Thu, 13 Aug 2026 09:49:19 +1000

I knew there would come a moment when having the time to hand-write code will become more of a luxury rather than just “how things are done”. I wasn’t expecting that moment to arrive this week.

Thu, 13 Aug 2026 09:26:14 +1000

I remember a time when I use to get a thrill playing around with word processors: adding headers with my name and the date, or crafting up VBA scripts for running macros. Now, I just want to get in, do what I need to do, and get out as quickly as possible.

Thu, 13 Aug 2026 07:20:22 +1000

I should add a general disclaimer about everything that’s posted here: “this is probably just a me problem.”

Thu, 13 Aug 2026 07:11:44 +1000

I’ve grown a real scarcity mindset with my RSS feeds, keeping posts things later rather than reading about them now. The sad fact is that I just don’t see many people write posts frequently enough. Or, to be fair, posts that I would be interested in reading at the time. There’s the odd essay or polished piece or two, which is nice. But it makes for focused reading which I need to be in the right mind for.

Not sure how I could fix this. Either I need to subscribe to more feeds, or start subscribing to Mastodon/Bluesky feeds via RSS. But it would be nice to see is more people writing on their own blog about the day to day, rather than saving those posts for social media.

Wed, 12 Aug 2026 17:21:59 +1000

Code generation tools — the conventional ones, not AI agents — can be huge PR size amplifiers. I updated the version of gRPC tools, and every single generated file changed. Made up a good 80% of the review diff. At least take the version number out of the generated file. Who’s going to need that?

Tue, 11 Aug 2026 15:01:41 +1000

Had to do some changes to NATS streams and consumers today, and learnt a few things from doing so.

First, it is possible to change the subject filters of a NATS consumer while it’s being used. That is, a client could be using it to pull messages, and it doesn’t need to reconnect or anything to get the effect of the consumer change. It will continue to receive messages, so long as they match the new subject filter.

Second, if you’re using the NATS CLI to change the filter, in order to specify multiple subject filters, include multiple --filter switches:

$ dc-nats consumer edit \
        --filter='topic.one' \
        --filter='topic.two' \
        --filter='topic.reset.>' my-stream my-consumer

Specifying a single --filter value with subjects separated by commas will not work. I’m not entirely sure what will happen but I suspect that it will create a subject filter that includes the commas in some way.

Tue, 11 Aug 2026 08:43:45 +1000

The most common integer type in Go is int, and 90% of the time, it should be the one to use. There are some QoL issues with using uint: values show up as hex in the debugger, linter complains about overflows. A 64 bit unsigned int is overkill for a retry count that’s likely to be less than 10.

Mon, 10 Aug 2026 08:42:24 +1000

We all had a laugh when we learnt that Accenture employees use AI agents to convert PDFs to Markdown. But just now, I had to get a Slack AI agent to convert a Canvas to Markdown, so I can pass it to another agent. It’s clear we need Markdown export support across the board now. The agents demand it.

On Doomerism

Mon, 10 Aug 2026 08:24:35 +1000

Terence Eden blogs about doomerism:

Are there terrible things which need changing? Absolutely! Are politicians doing everything that you specifically want? Probably not! Are incremental changes happening if you bother looking for the evidence? Yes. […] I guess a nuanced or positive take doesn’t get you as much social media engagement as a childish howl of rage.

It occurred to me that any talk of things getting better falls into that trough of nuance that always comes up in online discourse. Things take time to improve, and never in a straight line. It’s usually harder to find evidence of it happening, and it usually comes with a bunch of caveats.

Much easier to be a doomer about things. It’s a cheap thrill of attention that requires no effort or nuance on your part. It’s also cathartic, giving you the sense that you’re doing something about it when really, it’s probably not to change anything in the grand scheme of things.

Enticing, but ultimately a deal with the devil. I think every doomest post you make slowly adds to your fatalistic outlook that things will never improve. Why would you post something like that if a bit of you, maybe just a tiny bit, didn’t believe it to be so? Are you willing to wrestle with the cognitive dissonance that comes from that?

Sun, 09 Aug 2026 17:34:55 +1000

Was on a walk when I got stopped by this magpie, who flew in front of me and looked up expectedly. I’m guessing it mistook me for someone else.

Sun, 09 Aug 2026 09:24:35 +1000

TIL about Game DevOps: the idea of merging features to main without checking that it will break anything, cutting release branches, stabilising those, and merging the fixes back to main, which will most likely be unstable again by the next batch of merged features. The idea is that there are so many people working on the codebase that there’s no way to rely on CI/CD to ensure that main will be stable after merges. There’s just not enough hours in the day for CI/CD to run. So you don’t promise stability there. You spend your effort stabilising the release branches.

This is a complete reversal of the principals I learnt in my career, which is that main must always be production ready. Usually to allow for a release to be cut at any time (like, for example, patching an issue in prod), but also because no-one likes to checkout a broken branch. I should probably ask my colleagues that had experience in the game industry about this, and see how well it worked for them.

Follow up

Talked to my contact about this. He says that this is indeed a thing, and that main would usually be unstable because CI/CD runs would take too long. The reason, which I didn’t consider, is that the entire team is working towards a single artefact and it’s difficult splitting it out into separate repositories with their own CI/CD. But he did make the sense that it hated this form of working. It got so bad that one of the senior devs would need to spend every morning just stabilising main.