UP | HOME

On thinking machines

Something I've been thinking a lot about is, what makes a computer useful for thinking? Specifically, I want a machine that is set up to write exploratory code1 and write in my notebook. Some characteristics of a good thinking machine are that it should

  1. Have a decent battery life. I'd say at least 4 hours, but really, 8-12 hours is the sweet spot.
  2. Be quite portable. I really like my old Lenovo N100 (though it fails on the next point), but ideally for a laptop it would be a 12.5-14" screen. I want to throw it in a small bag (e.g. a Mk VII gas mask bag2) and not worry about it.
  3. It must be easy to type on. I liked my old Acer C720 (which I ran NixOS on), but the keyboard being a Chromebook keyboard was garbage. Likewise, the N100's keyboard was cramped. The Thinkpads generally have the best portable keyboard experience.
  4. It must be relatively durable. It's meant to be taken into the world and used, so it should stand up to some real-world use. Call it street computability, if you will.
  5. For right now, at least, I want to do emacs and common lisp on this machine. This puts a qualitative limit on the systems that can be used. While I would love a uLisp handheld (e.g. built on the PicoCalc3), that's not realistic right now.
  6. It should be inexpensive: this system is an additional system on top of my "EDC" system. I'm not willing to pay more than about $300 based on current finances for this.
  7. It's an ancilliary system, so it needs to quiet: there should be only enough software on there to write code, docs, and notes; and also to publish said writing. We'll call this attribute focus. It really means, this system is well-suited to being focused only on these tasks. There shouldn't be notifications or pings or what have you.

With this in mind, let's consider some examples:

Let's score them on our systems, with 0 being no, 1 being maybe, and 2 being a yes. A score of zero in any category will effectively disqualify a machine for consideration.

System Batt Portability KBD Durability Perf. Cost Focus Score
titan 0 2 1 1 2 0 0 4
FW12 1 2 1 2 2 0 2 10
imladris 2 1 2 2 2 2 2 13
hosaka 2 2 1 2 2 2 2 13

The clear winners heere are imladris and hosaka are the clear winners. Which… echoes my experience. A perfect score is probably impossible, as can be seen in the difference between the two: the uConsole is far more portable at the expense of a less pleasant (but perfectly usable) keyboard.

I'll say, roughly speaking, it's a 50-40-10 split (hosaka, imladris, titan) between contributing to my org site. I've gotten a ton of mileage from hosaka; it's pretty much my companion computer at this point. Sometimes the extra real estate (screen and keyboard) of imladris is worth it.

Footnotes:

1

Particularly in Common Lisp and Arduino/embedded C++ in service to building uLisp systems, or to write illustrative sketches for books I'm reading.

2

Also known as the Indiana Jones bag. It's a long story, but I found this out after getting one, and then watching Indiana Jones with the family over Christmas.

3

Work is ongoing here, and I'm sure the collapse computing tag in my roam notebook will capture some of the relevant ideas.

4

I specifically got it to be able to work with local LLMs, so it ended up being an investment in a minimum capability.

5

Something I should write about later. Essentially, I have a group of friends I swap hardware with when we realize we're not using a particular device and someone wants it.

Author: kyle (kyle@imladris)

Date: 2025-04-29 Tue 21:41

Emacs 30.1 (Org mode 9.7.11)

Validate