Hacker News: ks2048

New comment by ks2048 in "Low-Tech Ceramic Water Filter"

ks2048 — Mon, 17 Aug 2026 04:43:00 +0000

I'm in Antigua, Guatemala where the embedded video is from. As the video says, it may be "low tech", but it's not like something you can do very easily for yourself. Ecofiltro is everywhere here. But, it is more expensive then I would have thought - maybe $30 for a clay pot and $50 for it's container.

New comment by ks2048 in "Strait of Hormuz Live Traffic Tracking"

ks2048 — Sat, 15 Aug 2026 14:06:35 +0000

This like when a company has a status page, but the status page is in failing state.

New comment by ks2048 in "The case for overhauling American science"

ks2048 — Sat, 15 Aug 2026 00:27:39 +0000

> live in literal castles.

lol. For example?

New comment by ks2048 in "World Train Map – 1247 train routes around the world"

ks2048 — Thu, 13 Aug 2026 21:22:41 +0000

I called it AI slop because the first thing I looked for (Tren Maya) is missing half it tracks. Maybe that's not AI's fault, but that was my first impression.

https://worldtrainmap.com/?route=tren-maya

https://www.openstreetmap.org/relation/13196595#map=7/19.237...

New comment by ks2048 in "Mistral OCR 4.1"

ks2048 — Thu, 13 Aug 2026 20:57:09 +0000

Does anyone know a site that lets you browse examples of input / output pairs?, particularly with layout analysis (bounding boxes of figures, tables, etc).

New comment by ks2048 in "World Train Map – 1247 train routes around the world"

ks2048 — Tue, 11 Aug 2026 02:44:45 +0000

Whatever this AI slop is.. the better way would be to use OSM data. If you want more “curated” - there should be corresponding metadata in OSM.

New comment by ks2048 in "World Train Map – 1247 train routes around the world"

ks2048 — Tue, 11 Aug 2026 02:41:33 +0000

One is AI slop, one is direct OSM data

New comment by ks2048 in "World Train Map – 1247 train routes around the world"

ks2048 — Tue, 11 Aug 2026 02:34:07 +0000

How is it comprehensive? Did someone just say “be comprehensive” to their agent?

Quick Look in Mexico and it’s not good.

New comment by ks2048 in "World Train Map – 1247 train routes around the world"

ks2048 — Tue, 11 Aug 2026 02:32:55 +0000

AI lets people produce things that look authoritative, but are crap. Tiresome.

At first glance, https://www.openrailwaymap.org seems much better

New comment by ks2048 in "Depression has tripled in the last 15 years. Arthur Brooks about the cause"

ks2048 — Sun, 09 Aug 2026 00:41:31 +0000

To put it more succinctly: the world is changing too quickly for our biology to adapt.

New comment by ks2048 in "I am retiring from fulltime writing (& pseudonymity) to launch Guardian Angel"

ks2048 — Sun, 09 Aug 2026 00:39:10 +0000

My point was the opposite - I don't think this "Guardian Angel" product will be worth anything.

New comment by ks2048 in "The Unreasonable Effectiveness of Mathematics in the Natural Sciences [pdf] (1960)"

ks2048 — Sun, 09 Aug 2026 00:18:02 +0000

Never seemed unreasonable to me - math is “rule-based” and so is nature. The later is perhaps mysterious, but Occom’s Razor makes it more likely than lawless.

New comment by ks2048 in "I am retiring from fulltime writing (& pseudonymity) to launch Guardian Angel"

ks2048 — Wed, 05 Aug 2026 00:35:16 +0000

You’re assuming this $1000/mo will buy you something useful.

New comment by ks2048 in "Dates That Don't Exist (2015)"

ks2048 — Tue, 04 Aug 2026 19:56:30 +0000

I've been writing some programs to deal with the Maya Calendars, where it is an issue.

On some level, it doesn't matter if an inscription corresponds to "August 31, 83 CE" or "September 2, 83 CE" (a date potentially inscribed at the site Takalik Abaj in Proleptic Gregorian or Julian), but it's nice to try to get the details correct. And if you want to correlate with other events in history (especially if you can find astronomical event data from Europe), you need to be at least consistent.

Plus, it's just fun to think about these events from long ago in terms (dates) we understand.

New comment by ks2048 in "Dates That Don't Exist (2015)"

ks2048 — Tue, 04 Aug 2026 19:55:39 +0000

I would think of it as Calendar A uses Calandar B until time point T and then uses Calendar C. These dates exist in Calendars B and C, but do not exist in Calendar A.

New comment by ks2048 in "Scanning 7.6 Petabytes of HuggingFace Training Data for Secrets"

ks2048 — Sat, 01 Aug 2026 22:46:55 +0000

I think "7.6 PB" is more informative than "4.4x Empire State Building heights worth of DVDs", but that's just me.

New comment by ks2048 in "Show HN: Local text, image, video, music and 3D from one CLI, no Python"

ks2048 — Fri, 31 Jul 2026 00:17:22 +0000

At least add “models”. Doing “music” on your local machine can mean many things.

New comment by ks2048 in "Show HN: Local text, image, video, music and 3D from one CLI, no Python"

ks2048 — Thu, 30 Jul 2026 23:27:05 +0000

FYI, the title makes no sense in isolation.

New comment by ks2048 in "LearnVector – Andrew Ng's AI company building one‑to‑one learning experiences"

ks2048 — Wed, 29 Jul 2026 04:12:16 +0000

“Just one more AI company, I swear…” - last words of an AI-company-junkie.

New comment by ks2048 in "LearnVector – Andrew Ng's AI company building one‑to‑one learning experiences"

ks2048 — Wed, 29 Jul 2026 04:08:18 +0000

My understanding (in hindsight) is that success of Duolingo had nothing to do with “crowdsourcing” - is that correct? (Beyond maybe stories for fundraising).

It doesn’t seem to have a lot to say about “ed tech” - popularity of apps seems disconnected to popularity