Hacker News: ahartmetz

New comment by ahartmetz in "KPMG's AI report turns into a demo of AI hallucinations"

ahartmetz — Mon, 15 Jun 2026 09:14:23 +0000

Nah, it's useful but needs to be treated with suspicion due to persistent issues. Very persistent so far, not fixed or even improved much by new models, which is my point.

New comment by ahartmetz in "How to Earn a Billion Dollars"

ahartmetz — Sun, 14 Jun 2026 15:00:15 +0000

I mean how much you need to "take" / earn from each other citizen to accumulate the billion.

New comment by ahartmetz in "How to earn a billion dollars"

ahartmetz — Sun, 14 Jun 2026 12:31:44 +0000

You can exacerbate a problem (which Airbnb is doing) without being its major cause. Doing that is still bad.

New comment by ahartmetz in "How to earn a billion dollars"

ahartmetz — Sun, 14 Jun 2026 12:28:16 +0000

Yup. Billionaire - doable without too much ethical compromise (e.g early Google). The thing is that a billion isn't much per "first world" citizen if that is roughly your market. Many-many-billionaire - would need some good example, I can only think of bad ones. The SAP guys maybe? But they aren't exactly the richest. The only bad things I've heard about SAP are "making clunky software" and "charging too much money".

New comment by ahartmetz in "How to earn a billion dollars"

ahartmetz — Sun, 14 Jun 2026 12:17:42 +0000

Unfortunately, there was plenty of not understanding exponentials on display during Covid, including from politicians, journalists and other public figures.

New comment by ahartmetz in "Codex for open source"

ahartmetz — Sun, 14 Jun 2026 11:49:19 +0000

I want my license fee from my software that they trained their models on.

(Actually I don't, I want their stuff as Free Software and I mean everything, training data, pipelines and all)

New comment by ahartmetz in "A new era for software testing"

ahartmetz — Sat, 13 Jun 2026 20:27:41 +0000

IME there are these levels of tests:

- If you call the setter, the getter returns the same value - these are kinda bullshit and would be caught by the next level anyway

- Testing basic normal use

- Testing known difficulties of the implementation

- Exhaustive or randomized (if necessary) testing of the state space, ~= property-based testing

I expect AI to have very different levels of ability for these, not necessarily in strictly descending order as listed.

New comment by ahartmetz in "Leaving Mozilla"

ahartmetz — Sat, 13 Jun 2026 15:16:59 +0000

True, but the bad outcomes are mostly a thing of the past. PipeWire has better reliability, lower CPU usage, and lower latency.

New comment by ahartmetz in "Swift at Apple: Migrating the TrueType hinting interpreter"

ahartmetz — Sat, 13 Jun 2026 12:00:31 +0000

All 150 kloc in six months by two people? Actually, it sounds like way too much code for the task unless 70+% of it is tests.

New comment by ahartmetz in "Swift at Apple: Migrating the TrueType hinting interpreter"

ahartmetz — Sat, 13 Jun 2026 11:46:36 +0000

Lower DPI is quite okay with good font rendering (FreeType, slight hinting, subpixel rendering). I have both kinds of devices.

New comment by ahartmetz in "KPMG's AI report turns into a demo of AI hallucinations"

ahartmetz — Sat, 13 Jun 2026 11:33:10 +0000

Every report about AI problems has this comment. Just one more model, bro.

New comment by ahartmetz in "Leaving Mozilla"

ahartmetz — Sat, 13 Jun 2026 08:58:18 +0000

The current one is PipeWire (it's much better)

New comment by ahartmetz in "I Am Not a Reverse Centaur"

ahartmetz — Sat, 13 Jun 2026 01:15:31 +0000

I've done the same for a fix in the Gold linker, which is now obsolete due to even faster linkers being available. Shout out to Ian Lance Taylor, his behavior as the maintainer was exemplary: very gracious and very responsive.

New comment by ahartmetz in "Renault: Electric motors with no rare earths"

ahartmetz — Sat, 13 Jun 2026 00:53:43 +0000

After watching a Munro video about it, I see your point. In the motor shown, the rotor gets its magnetic field simply by inducing a current and a field in it in reaction to the stator's field. There are no electromagnets in the rotor like I expected. In that case, I'm not sure either... I'd say more likely than not but it's complicated since the stator basically needs to induce a field and at the same time recover energy from the field that comes back from the rotor. I would further guess that the phase shift between the two components makes it possible to treat them separately.

Previous comment: Don't see why not - the "field" coils (the ones that replace the permanent magnets) need to be energized, which can initially come from the batteries if necessary.

New comment by ahartmetz in "Electric motors with no rare earths"

ahartmetz — Sat, 13 Jun 2026 00:50:44 +0000

Because it's the discontinuities in the commutator where the sparks fly (with much help from self-induction of the motor's coils) and erode the ring and brushes.

New comment by ahartmetz in "I Am Not a Reverse Centaur"

ahartmetz — Fri, 12 Jun 2026 21:47:56 +0000

If you already know it's not good enough, you can just say so by calling it a proof of concept or hack to demonstrate what needs to be done. Such code is often very useful when writing the real fix.

New comment by ahartmetz in "Ryanair dark UX patterns summer 2026 refresher"

ahartmetz — Fri, 12 Jun 2026 15:01:31 +0000

IIRC Aer Lingus could already be described like that 10-15 years ago. I remember it as cheap but decent.

New comment by ahartmetz in "Linux latency measurements and compositor tuning [KWin Wayland]"

ahartmetz — Thu, 11 Jun 2026 19:58:29 +0000

I chose my monitor very specifically for low latency, good color reproduction and "high" (as in the screen is higher than 16:9) aspect ratio - in about 2012! and I'm still quite happy with it. 4K is incoming, but only if it's better in all criteria, which became available fairly recently.

A lot of people still seem to buy their monitors without considering latency, which majorly sucks. It feels like the years between power supply efficiency making a difference and the introduction of 80+ efficiency standards.

New comment by ahartmetz in "Linux latency measurements and compositor tuning [KWin Wayland]"

ahartmetz — Thu, 11 Jun 2026 11:22:56 +0000

These latency numbers all look terrible to me tbh. I got better latency numbers (a little over 2 frames in KWrite, close to the theoretical minimum with X11 compositing) with an AMD GPU w/ open drivers, KWin, X11 and compositing enabled a couple of years ago. Measured with high fps recording on a phone camera filming keyboard and screen. Both Wayland and KWin changes should have improved the situation since then. I suspect the games, the screen and / or the graphics drivers which are presumably "review optimized" for frame rate over latency.

Actionable advice from the article: Using Wayland natively from Wine does seem to help, especially against latency jitter / for consistent frame pacing, which is a typical improvement for Wayland.

New comment by ahartmetz in "Raspberry Pi 5 – 16GB RAM"

ahartmetz — Wed, 10 Jun 2026 21:57:03 +0000

They actually have a factory in the UK, one that previously made Sony TVs IIRC.