Hacker News Reader

It's tailscale or local LAN. syncthing allows you to specify multiple addresses for the same peer

> tailscale or home routers network

cl3mischon Fitting Neural Textures and PBR Material Maps with ES (No Backprop)

> no backprop, no autodiff, no training framework

OP says it like it's a good thing, but doesn't explain why. To not rely on Pytorch? What's the benefit?

> or maybe this is just a retelling of stoicism and endless levels of historical philosophy

I think what OP describes is much more Epicureanism than Stoicism? Stoicism's goal is "living a well-reasoned life" but based on a global logic system and reasoning about the world. Epicureanism is much more about living well enough in your little bubble (which I highly appreciate).

cl3mischon To become a better writer, read as much as you can

But are these game devs designing the gameplay loop? Or are they working on backend, like the engine?

I could imagine a (solo) game dev not self-describe as a gamer, but they must have some intuition about what makes (video/board) games fun and engaging.

I can't imagine you can obtain that without ever playing games yourself?

cl3mischon Show HN: Rex, a parallel functional language for scientific workflows

Yeah but pretty much every name will be taken by something? As long as it's not too close to your new project, I think it doesn't matter that much.

cl3mischon Show HN: Rex, a parallel functional language for scientific workflows

I expected a reference to Dex, but they seem to be unrelated?

https://github.com/google-research/dex-lang

cl3mischon What happens when an LLM never sees material beyond fifth grade?

> Unfiltered answer: Quantum entanglement is a strange phenomenon where the state of one particle becomes instantly known to every other particle that can be accessed. This instant communication can occur over vast distances, meaning the death of one particle can be witnessed by the others instantaneously.

Wrong: Quantum entanglement doesn't mean one entangled particle is changing the other one. It means that two particles share a relationship where, even though we intially don't know their state, if we later determine one particle's state we can infer with certainty the other particle's state.

This has been common and popular misconception long before LLMs. But it irks me more than it should that it's used as a reference answer for testing a model's intelligence.

cl3mischon Why does Opus 5 feel worse to work with?

But wouldn't you have to know the exact context before this token in order to verify the watermark? I.e. a paragraph wouldn't be sufficient; you would need the system prompt, previous prompts, and even hidden thinking?

cl3mischon Accelerating GPT-5.6 Sol Ultrafast

> In fact, one could almost imagine it baked into the assistant/harness

...it already is, via thinking/effort. That wouldn't be possible if a LLM at usable quality wasn't fast enough to allow at least some amount of "thinking" (i.e. hidden text generation).

But your point still stands: we could get massive quality gains by allowing even more thinking by default, if it was fast enough.

cl3mischon Codex in ChatGPT desktop app for Linux is now in preview

I prefer the CLI myself, but I could imagine:

- nicer diff rendering, and directly mentioning parts of a diff

- Latex rendering (pretty annoying in the CLI, you have to actively tell Codex it's running in a terminal and replace Latex syntax with UTF-8 glyphs)

cl3mischon The hardest working font in Manhattan (2025)

https://archive.ph/mYRQX

(the site seems to eagerly load lots of large images, and a 500MB video)

cl3mischon Incentives are for losers

> Up close, when you can see what they are, it becomes clear that only a few of these ladder-toppers are indeed wise and cool. More often, they’re sniveling little twerps. They’re spiritually chinless. They have no convictions, no passions, no purpose other than ladder-climbing.

I like the article overall but this part seems overly uncharitable. I think the next step of your development would be to have your own incentives without looking down on other's who you think lack their own. You should be able to know who you are without shitting on others.

cl3mischon Triton: DirectX 11 Driver for QEMU

I was expecting DirectX in triton-lang https://github.com/triton-lang/triton

cl3mischon Minecraft Java raises recommended memory to 16GB ahead of Vulkan transition

Wasn't that the Bedrock edition? Or "Minecraft Education Edition"? An alternative C++ version, anyway.

I can't imagine Java Minecraft ever to have run well on a mid-range Raspi.

cl3mischon Measuring Input Latency on Linux: X11 vs. Wayland, VRR, and DXVK

Inspired by my own comment I looked into it again and came up with this combination in the game's "launch options", which feels much better:

    DXVK_CONFIG="dxgi.syncInterval = 0; dxgi.maxFrameLatency = 1; dxgi.maxFrameRate = 60" gamescope -f --immediate-flips -- %command%

I didn't know about `DXVK_CONFIG` before. It's the DirectX-Vulkan translation layer with its own framebuffer and latency.

I also manually disabled vsync in the game's `local.cfg` XML file.

I love where "gaming on Linux" is going. But I have the feeling the hype on Youtube etc. is in part created by content creators, and doesn't faithfully represent the actual situation (surprise!).

cl3mischon Teardown: A Generic 7-Port USB 3.0 Hub That Wasn't

So it's not a "hub", but a box around extension cables?

cl3mischon Measuring Input Latency on Linux: X11 vs. Wayland, VRR, and DXVK

This hits a very soft spot for me. My beloved Spelunky 2 has huge input lag "on Linux", which makes it unplayable for me. It's pretty much the reason I keep my Windows 10 install around. I could not come up with a configuration on Linux that avoids the lag.

It's extra infuriating that online sources report the game running flawlessly but seem to ignore input lag.

cl3mischon Show HN: Analog Watch

It seems to not support 24h? When I see e.g. 17:35 I want to type 17:35 instead of 5:35. Probably because I very seldomly see 5:35 in real life! Interesting observation by itself, but slightly annoying that I have to convert to 12h for this game.

cl3mischon Programmers need to start meditating

> Tools are supposed to be designed around humans

This is a common thing to say, but when during the development of human civilization has this actually been the case? Is agriculture designed around humans more than hunting/gathering? Is industrialized work more designed around humans than agrarian society?

I don't mean to sound pessimistic or technocratic; quite the contrary. But I think we shouldn't project our desire for equanimity onto romantized versions of civilization.

cl3mischon Africans Are Turning to Starlink

Is this allowed in Starlink's terms of service? Building a commercial service by selling a part of your antenna connection?

cl3mischon Night Witches – all-female Soviet aviator regiment WW2

From the depths of hell in silence

Cast their spells, explosive violence

Russian night time flight perfected

Flawless vision, undetected

(Sabaton)

cl3mischon Instead of banning AI, I made a classroom contract with my students

Moderation in recreational drugs, which definitely rules out fentanyl?

But yeah, what's moderation and what's excessive is subjective.

cl3mischon Immich 3.0

I can only comment about battery life. It's proportinal to how much tailscale is really being used: If you use tailscale with an "exit node", i.e. all traffic is routed through it and it's working continously, it drains battery. If it's only used for services on your tailnet, e.g. Immich, the impact will be very small.

cl3mischon Have you restarted your computer this week?

But that doesn't give you kernel updates, being userspace only?

cl3mischon Librepods: AirPods liberated

PR #655 of which project? I thought you meant Wireplumber but that's not it: https://gitlab.freedesktop.org/pipewire/wireplumber/-/merge_...

Edit: You meant Librepods. Apparently I managed to forget what we were commenting on in the first place.

cl3mischon Mojo programming language will become open-source soon

What exactly? I already quoted what I could find (only very general statements).

Ah, seems to be here: https://mojolang.org/docs/faq/#will-mojo-be-open-sourced

> We have committed to open-sourcing Mojo in Fall 2026.

cl3mischon Librepods: AirPods liberated

Interesting! It reads to me like Apple advertises AAC-ELD only for Facetime though?

And the alternative, newer "Bluetooth high quality recording" I mentioned in sibling comment is not recommended for real-time communication due to increased latency.

So the situation seems to be more complex than I thought. But I stand by my original sentiment: I dislike the prevalence of bad-sounding Airpods in meetings.

cl3mischon Librepods: AirPods liberated

They use hardware tricks (beamshaping) and software tricks (audio processing) to get as much out of them as possible, and they do sound better than generic cheap Bluetooth earbuds, but the Bluetooth connection is still the bottleneck.

They did announce a special "high quality recording mode" for Airpods in iOS 26 (which I didn't know, so thanks) but I couldn't find much except their developer announcement. It's not documented and I don't have the feeling it's being used.

cl3mischon Librepods: AirPods liberated

> Note that this is the default for better quality (Bluetooth spec, not Linux).

Yeah, I have the impression most people assume Apple magically solved sound/mic quality in bidirectional mode. Or they just don't care. In any case, I'm unhappy about how much Airpods are used for meetings. They sounds so much worse than any wired mic! Or your laptop's mic, as long as you're not using the speakers.

cl3mischon Working around dragons with the Lemote Yeeloong laptop and OpenBSD

His long list of conditions is an interesting read. It comes off as arrogant in places, but I guess if you really travel a lot and have unusual requirements it's best for everyone to be upfront about it!