Hacker News: lukaslalinsky

New comment by lukaslalinsky in "Btrfs/ZFS/bcachefs under workloads classic benchmarks skip"

lukaslalinsky — Sun, 20 Sep 2026 05:46:02 +0000

I've been using ZFS on my Linux servers for years and have pretty good experience with it. I've not been following Linux development too much. Is bcachefs usable/stable/reliable enough to replace it?

New comment by lukaslalinsky in "What Zig felt like, coming from Rust"

lukaslalinsky — Sun, 20 Sep 2026 05:20:55 +0000

The main advantage of Zig for me and many other programmers, is the simple mental model you need to keep while reading the code. All control flow is expressed by keywords, never symbols. That along makes code much more readable. There is nothing like Rust macros, things can't do "magic. Zig comptime are regular functions, so you can clearly see and follow the trail. There are no exceptions, no recoverable panics, all code will run linearly as you see it written. Error handling in Zig is close to perfect, in my view. Obviously, Zig is not a language to use for high-level jobs. It makes no sense to write CRUD web app in Zig, you will be fighting the resource management for no good reason. For me personally, I feel that I'm too stupid to use Rust effectively. I started moving away from C++ purely because I don't want to deal with exceptions. And Rust is just a more complex C++ with memory safety.

New comment by lukaslalinsky in "Europe's summer drought is so extreme that desertification is a growing threat"

lukaslalinsky — Mon, 31 Aug 2026 08:02:49 +0000

The concept of winter pretty much stopped existing where I live. It's just colder autumn. The ground doesn't freeze. Snow doesn't accumulate. No amount of meandering is going to compensate for the fact that the mountains do not hold water. Something needs to feed the springs and rivers. And the source is gone. Europe is shifting into a different climate, where water primarily comes as rainfall, causing floods. I imagine it's similar to Asian countries, and we will have to learn to deal with that.

New comment by lukaslalinsky in "Europe's summer drought is so extreme that desertification is a growing threat"

lukaslalinsky — Sun, 30 Aug 2026 16:16:39 +0000

I had a similar observation this year. In 2015, I was in California, was doing some hiking on the weekends, and I found it very unnatural how all the grasses were dry, trails dusty, on the hills/mountains. I came back to Slovakia, and I was happy to see the green grass. Fast-forward to 2026, and our grasses are equally dry and trails are equally dusty. Water management seems to be more important than ever, because this is only going to get worse. Maybe we need to build massive reservoirs, to retain the flood water somehow, if we want to keep living here in the next 100 years.

New comment by lukaslalinsky in "Europe's summer drought is so extreme that desertification is a growing threat"

lukaslalinsky — Sun, 30 Aug 2026 16:07:48 +0000

People are not scared of individual droughts. They are scared of the patterns. Here in Slovakia, the last real winter with proper snow depths we had was in 2018, if I remember correctly. There is pretty much no snow anymore. All the water comes in form of heavy rainfall, causing massive floods, and the water goes straight to the sea. Nature depends on the water accumulating on the mountains during the winter, and that just doesn't seem to happen, at least in this part of Europe.

New comment by lukaslalinsky in "Zig’s Io.Threaded is neat"

lukaslalinsky — Sat, 22 Aug 2026 18:15:37 +0000

Programming asynchronous code is hard. Execution is split up, you have extra state machine to handle, extra resources to manage that need to survive across async boundaries. And by asynchronous, I mean either callbacks or manually checked completions from some event system. Modern languages hide that under async/await, which is much better from DX perspective, but still leads to split in the ecosystem, and it's hard to optimize across async boundaries.

This is an example why Go is such a successful language, in my view. It hides the async complexity and allows you to pretend you have a simple continuous thread of execution. It's just cheaper than system threads. Imagine if operating system threads and the blocking syscalls were this efficient.

I've spent the last year building a similar runtime for Zig. Purely because I want my applications to forget they are using things like io_uring in the background.

New comment by lukaslalinsky in "Zig’s Io.Threaded is neat"

lukaslalinsky — Sat, 22 Aug 2026 13:20:40 +0000

If you are comparing to overlapped I/O, you should use io_uring as reference and cancellation in io_uring is fairly simple.

New comment by lukaslalinsky in "Mojo is now open source"

lukaslalinsky — Thu, 20 Aug 2026 16:04:31 +0000

One problem is that a large part of Europe doesn't understand the concept of public domain. I'm not legally allowed to not own my creation, except by transferring it to someone else. That's why licences like CC0 exist.

New comment by lukaslalinsky in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

lukaslalinsky — Tue, 11 Aug 2026 06:48:23 +0000

I run this on a side of the Claude Pro subscription that I use for other purposes. My main motivation was root cause analysis of production issues. I have a solo project and unfortunately my mental state has been degrading over the last years. I would avoid looking at production issues, because I didn't have the energy to focus on the investigation. So I automated this, setup the loop, setup metrics/logs access for Claude to use and now whenever something goes bad, I have a single report that I can act on easily, and if I don't, it will ping me in a way that's not spammy like automated alerts. But I'm finding more uses for it.

New comment by lukaslalinsky in "Muse Glimmer: 30B-parameter model optimized for always-on local agent workflows"

lukaslalinsky — Mon, 10 Aug 2026 20:35:08 +0000

This is is already possible with Claude Code. I use a setup where I have one instance monitoring a local queue, I have a web app for receiving webhooks from various sources and pushing them to the queue. Plus email for things that don't have webhooks. That instance then decides what to do with each input, sometimes it can spawn additional agent to investigate/prepare, sometimes it creates a ticket assigned to me and then waits for me input. All of that just uses the monitoring tools built into CC. The dispatcher loop doesn't need to be extremely smart, so I might experiment replacing it with a local model like this.

New comment by lukaslalinsky in "LLMs reward expertise"

lukaslalinsky — Tue, 04 Aug 2026 05:38:11 +0000

Of course they do. They have such a huge parameter maps. You need to be able to guide it through the map, so it starts making the right connections. Even in the Sonnet 3.7 days, it became clear to me, that if I have want efficient code out of it, I need to really take care of the context. If I just let it research a problem, it will mess up most of the time. If I tell it to study A, B, C and then present problem D, it will solve it perfectly. And it's true even with the current top models.

New comment by lukaslalinsky in "Show HN: Getting GLM 5.2 running on my slow computer"

lukaslalinsky — Fri, 10 Jul 2026 10:12:46 +0000

That was their headline for Opus 4.8, I guess the invested into some post-training to get them to write this, and I like it, it's a great way to identify AI posts from Claude.

https://www.anthropic.com/news/claude-opus-4-8

New comment by lukaslalinsky in "Fable 5 wrote a Windows kernel in 38 minutes"

lukaslalinsky — Wed, 08 Jul 2026 05:08:21 +0000

God I hate reading these LLM-written articles. It also feels like it was one-shotted by Claude, not just the kernel.

New comment by lukaslalinsky in "No LLM Code in Dependencies"

lukaslalinsky — Sun, 05 Jul 2026 05:31:41 +0000

If I had coworkers with the ability to program like Sonnet 3.7 did, I'd be extremely happy. In reality, I can probably count 1-3 people like that, out of many over the years. Honestly, pair programming with Opus is better than any pair programming I had with a human, especially at understanding the reasons behind my designs.

New comment by lukaslalinsky in "Will It Mythos?"

lukaslalinsky — Tue, 23 Jun 2026 15:25:07 +0000

As much as I hate to say this, I think it is an user error. Fable is very to the point, much more so than any other Anthropic model. I found it to be cheaper to use Fable, than using Opus for same task, but in order to achieve that, it needs to be given a targeted task.

New comment by lukaslalinsky in "Will It Mythos?"

lukaslalinsky — Tue, 23 Jun 2026 15:21:18 +0000

My hope is that Opus 5 will be released soon, basically a rebranded Fable.

New comment by lukaslalinsky in "GLM 5.2 vs. Opus"

lukaslalinsky — Mon, 22 Jun 2026 20:04:16 +0000

I'll have to try it. I was using earlier GLM models, incluing 5.1, and was always disappointed.

New comment by lukaslalinsky in "GLM 5.2 vs. Opus"

lukaslalinsky — Mon, 22 Jun 2026 19:48:43 +0000

I considered Opus 4.5 to be the peak for a while. Opus 4.6 tended to over think, and generally get lost in thinking. I asked something and Claude Code would just spin for 15 minutes. And it was not the harness, if I changed the model to 4.5, it was fine again. So I skipped the following releases. I've been working with Opus 4.8 the last weeks and while I don't like how talkative it is, but it is fine to work with interactively. I've also used Fable for the few days it was available, and indeed, that was model worth using for my use case. To the point, but still very interactive.

New comment by lukaslalinsky in "Pledging another $400k to the Zig software foundation"

lukaslalinsky — Mon, 22 Jun 2026 15:07:02 +0000

Zig has multiple issues, but syntax is definitely not it. It might take a little bit of time getting used to, if you are coming from another language, but it's one of the most readable languages I've ever worked with.

New comment by lukaslalinsky in "GLM 5.2 vs. Opus"

lukaslalinsky — Mon, 22 Jun 2026 14:07:57 +0000

I was never able to get these models to collaborate with me the way Opus does. I'm probably an outliner, I don't one-shot projects, I don't vibe code. I basically use LLMs are if I was working with a coworker, fairly smart one, but with short memory and often missing the big picture. Sometimes I can delegate more, sometimes less, but I know I always have to stay on top of what's happening, because it WILL create mess when it hits something hard. With the Antropic models, this kind of cooperation is easy (with the exception of Opus 4.6, which was bad for some reason).