Skip to content

Instantly share code, notes, and snippets.

user

why is potassium called that in English and Potasio in Spanish, but it's Latin name is kalium and German uses the Latin word?

agent

The difference comes from two competing names that appeared when the element was discovered in the early 1800s, and different language groups adopted different ones.

1. The English name: potassium

The English word comes from potash, the material from which the element was first isolated in 1807 by the chemist Humphry Davy. ([Wikipedia][1])

  • Potash literally meant “pot ashes.”
@rma92
rma92 / edid.md
Last active February 18, 2026 20:02
GLKVM GLi.net Comet

These are EDIDs to use with the GLKVM (GLi.Net Comet Pro) with a 3440x1440 monitor.

Note that the Comet Pro can only do 3440x1440 up to 50 Hz, as it's limited to HDMI 2.0. The Crua 3440x1440 monitor I have does not work at 50Hz, so this isn't working.

"3440x1440@60Hz", makes 2560x1080 on Mac

00 ff ff ff ff ff ff 00 05 e3 63 29 4b 01 00 00
0a 17 01 03 80 43 1c 78 2a ca 95 a6 55 4e a1 26
0f 50 54 bf ef 00 d1 c0 b3 00 95 00 81 80 81 40
81 c0 01 01 01 01 cd 46 00 a0 a0 38 1f 40 30 20
@rma92
rma92 / chatgpt.md
Last active February 10, 2026 23:54

user

Which of these are the fastest for running an LLM, assuming the same VRAM:

  • Nvidia A100
  • Nvidia A40
  • Nvidia L40S
  • Nvidia GH200

agent

Fastest → slowest for running an LLM (inference/training), ignoring VRAM capacity differences:

agent: ChatGPT 5.2 Auto

user

For some reason, people seem to be funny about using electric screwdrivers / drills / screw guns while working with wiring devices in the US. Is this an accurate assessment? If so, any idea why? I've seen professional electricians use powered drivers - it saves a ton of time and saves stamina!

agent

Short answer: yes, that perception is accurate in the US, and it’s driven more by code interpretation, safety culture, liability, and training norms than by a hard technical prohibition.

Below is the breakdown.


agent: ChatGPT 5.2 Auto

user

How does the current Jeep Wagoneer compare to the Grand Wagoneer? The prices are similar, and I'm looking specifically at renting from National / Enterprise. I routinely see Grand Wagoneers there, and my friend rents them on a regular basis. The "Premium SUV AWD" class lists "Jeep Wagoneer L". Pay particular attention to any notes from Flyertalk.

agent

Here’s a practical, rental-focused breakdown of the Jeep Wagoneer vs Grand Wagoneer — especially relevant if you’re seeing “Premium SUV AWD – Jeep Wagoneer L” in National/Enterprise fleets and your friend often gets Grand Wagoneers. I’ll also include FlyerTalk real-world rental impressions where available.


Agent: ChatGPT 5.2 Pro Thinking

user

Being an experienced software and electrical engineer who also has a ton of experience on the project management, specification, and manual writing side, is it likely that I'm getting better "vibe coding" results than usual considering that I write extremely detailed and thorough specs that break things down into a way that a human could easily complete the various coding tasks?

agent

Yes—very likely.

“Vibe coding” quality is heavily driven by the quality of the problem formulation. If you routinely write specs that a competent human could execute with minimal ambiguity, you’re feeding the model the same kind of structure it needs to produce coherent code: clear intent, constraints, interfaces, and acceptance criteria. Your software + electrical engineering background plus PM/spec/manual-writing experience also tends to reduce two common failure modes: missing requirements and vague definitions.

Set up Podman Compose on Debian

Note: podman-compose uses rootless podman to do the compose. podman compose uses the socket and root daemon.

sudo apt install podman podman-compose podman-docker

systemctl --user enable --now podman.socket
systemctl --user status podman.socket

# confirm it works
@rma92
rma92 / claude-code-codex-file.md
Created February 5, 2026 18:59
Claude Code and Codex using a long instructions file.

user

Do Claude Code or Codex support a way to pass in a very long message from a file? For example, I want to provide the instructions to start building the application. It's several paragraphs, so I want to write it with vim, save it, then tell Claude or Codex to read the file and follow the instructions.

Does this work? Can you find anything online? Ideally try to cover both tools.

Approaches that seem reasonable are:

  • Writing the file, say, instructions.md or instructions.txt in the directory and telling the agent to read it and build what is described.
  • Pasting the contents of the file into the interactive prompt.

I'm open to other options. Please try to provide commentary for both tools, as we use both of them at my firm. Thanks for your help, you're the best.

@rma92
rma92 / Auto-saved 4eacaa96.xml
Created November 21, 2025 12:55
rma92 Windows 11 LTSC 2024 Lite with no components removed.
<?xml version="1.0" encoding="utf-8"?>
<Preset isAutoSaved="true" xmlns="urn:schemas-nliteos-com:pn.v1">
<Date>11/21/2025 12:42:09</Date>
<AppInfo>
<Version>2025.8.10552</Version>
<Licensed>Yes</Licensed>
<Protections>Yes</Protections>
<Host>Windows 10 EnterpriseS 21H2 x64 - 10.0.19044.6575 (en-US)</Host>
</AppInfo>
<ImageInfo>
@rma92
rma92 / Auto-saved c297fbaf.xml
Created November 21, 2025 11:40
Windows 10 LTSC 2021 rma92 Lite
<?xml version="1.0" encoding="utf-8"?>
<Preset isAutoSaved="true" xmlns="urn:schemas-nliteos-com:pn.v1">
<Date>09/12/2022 13:25:18</Date>
<AppInfo>
<Version>2.3.8.8890</Version>
<Licensed>No</Licensed>
<Protections>Yes</Protections>
<Host>Windows Server 2022 Standard x64 - 10.0.20348.405 (en-US)</Host>
</AppInfo>
<ImageInfo>