Hacker News: gct

New comment by gct in "Compiler-Assisted Floating-Point Error Analysis and Profiling with FPChecker"

gct — Wed, 01 Jul 2026 13:49:55 +0000

Is there actually video of the talk?

New comment by gct in "Starship V3"

gct — Wed, 13 May 2026 03:45:32 +0000

They'll put up thousands more starlinks and track every mobile device on the planet simultaneously, might as well have a homing beacon in your pocket.

New comment by gct in "A Declaration of the Independence of Cyberspace"

gct — Sun, 10 May 2026 18:59:41 +0000

We thought this was so punk back in the day.

New comment by gct in "Artemis II safely splashes down"

gct — Sat, 11 Apr 2026 00:56:17 +0000

Orbits do not work that way

New comment by gct in "Making geo joins faster with H3 indexes"

gct — Tue, 10 Feb 2026 02:22:53 +0000

I'd be curious to hear more about how you do the 2 -> 3 embedding there. In S2 it uses cartesian three space, but points are constrained to be unit magnitude. This has advantage and disadvantages obviously.

New comment by gct in "Making geo joins faster with H3 indexes"

gct — Sun, 08 Feb 2026 17:32:06 +0000

I'd like to hear more about the synthetic part of these three spaces, because S2 works exactly as you say, embedding the 2-sphere in three (cartesian) dimensions. S2 points are always three dimensional.

New comment by gct in "Making geo joins faster with H3 indexes"

gct — Sun, 08 Feb 2026 17:16:24 +0000

You can do this with [S2](https://s2geometry.io/) as well which has the very nice property that parent cells do indeed always contain their children, and sorting the cell ids puts them into in-order order.

New comment by gct in "Data centers in space makes no sense"

gct — Wed, 04 Feb 2026 00:25:36 +0000

Mars which is literally covered in poisonous perchlorate soil

New comment by gct in "Exe.dev"

gct — Sun, 28 Dec 2025 08:17:13 +0000

emacs!

New comment by gct in "How I Left YouTube"

gct — Thu, 25 Dec 2025 03:20:19 +0000

The Box is very frustrating, especially when there's no one handling the other things, yet you're still not allowed to do them because it'd make the wrong people look bad.

New comment by gct in "Leaving Intel"

gct — Sat, 06 Dec 2025 03:24:38 +0000

Managers can be lazy just like anyone.

New comment by gct in "Leaving Intel"

gct — Sat, 06 Dec 2025 02:29:07 +0000

At big tech you have to quantify your value like this regularly, so yeah everyone keeps track of the minutiae.

New comment by gct in "An individual can change an organization"

gct — Tue, 04 Nov 2025 15:54:05 +0000

This person admits he drove people to quit with his behavior, that goes wayyyyyy beyond trying to get the full details. I've worked with this type before and I can envision the cold sweats in the office as he comes around to hold people hostage explaining every micro-detail of their work until he understands it.

New comment by gct in "An individual can change an organization"

gct — Tue, 04 Nov 2025 15:52:30 +0000

I've worked with this particular kind of toxic personality, they think until they've personally internalized all the details, that its not "right", and will exhaust everyone doing so.

New comment by gct in "An individual can change an organization"

gct — Tue, 04 Nov 2025 04:30:06 +0000

One of those that thinks everyone owes them explaining everything to them in every detail or its not "right".

New comment by gct in "Charlie Kirk killed at event in Utah"

gct — Thu, 11 Sep 2025 02:09:30 +0000

They were also loose powder hand loaded weapons, you could fire three rounds a _minute_ if you were really skilled. Everyone in town had to store their powder in a (secure) communal location because it was, duh, an explosive.

New comment by gct in "Everything I know about good API design"

gct — Mon, 25 Aug 2025 00:00:13 +0000

Everyone's decided that writing regular software to run locally on a computer is the weird case and so it has to be called "local first".

New comment by gct in "Digg.com is back"

gct — Wed, 20 Aug 2025 17:35:08 +0000

Let's just start shifting the overton window: let's make all paid advertisement illegal y'all.

New comment by gct in "LLM codegen go brrr – Parallelization with Git worktrees and tmux"

gct — Thu, 29 May 2025 01:45:25 +0000

No wonder software is so slow today when we're this profligate. "let's run four AIs and ignore 3/4 of them!" ugh.

New comment by gct in "The FFT Strikes Back: An Efficient Alternative to Self-Attention"

gct — Wed, 26 Feb 2025 14:06:38 +0000

They usually have dedicated acceleration hardware, yes: https://www.ti.com/lit/an/sprabb6b/sprabb6b.pdf?ts=174057874...