David Huggins-Daines

Étudiant en géomatique, consultant en linguistique, développeur de logiciels libres (et autres). Piéton, cycliste et skieur. Ex-conseiller municipal et heureux de l'être. Posts randomly in French and/or English. Fan of cheap, durable, yet supposedly obsolete technology in all domains, with the notable exception of 3-pin ski bindings, those are bad and you should not use them.

Site web (professionnel): https://ecolingui.ca/
GitHub: https://github.com/dhdaines
Site Web (politique): https://dhd6.ca
Métal et ski de fond (par en bas): https://metalhead.club/@waxhammer

Last Notes

2026-09-28 18:17:55 UTC

RE: https://mastodon.energy/@Sustainable2050/117349913072181143
Basically every poll in every developed country everywhere 🫠
#note10lq…xt47

2026-09-01 12:34:17 UTC

- reply

but also, despite his association with Thatcher and Reagan, Mulroney was nonetheless a member of a totally different party than today's CPC.
Maybe because I came of age in the 90s but I think the Chrétien Liberals did a lot more austerity, with a lot more lasting negative effects, than the PCs ever could in the 80s.

2026-08-31 15:10:08 UTC

- reply

also for many things a filesystem really is the best representation, so for example instead of MBTiles which has lots of overhead, you can get most of the benefits of a structured, self-documenting format (as opposed to a Protobuf abomination which shall remain unnamed) by simply putting your tiles in a loop-mounted btrfs. I love when the answer to "what is the tech stack" is "there is no tech stack": https://github.com/hyperknot/openfreemap#what-is-the-tech-stack

2026-08-31 14:57:17 UTC

- reply

to be honest I have mixed feelings about the use of SQLite in these cases as the resulting files are much larger, for example here's a GPKG which is nearly *four times* larger than the (proprietary) ESRI FGDB format: https://www.donneesquebec.ca/recherche/fr/dataset/decoupages-administratifs

2026-08-30 13:40:13 UTC

- reply

the geospatial community seems to have come to this conclusion, see GeoPackage, MBTiles, etc
(It helps that Shapefiles literally have the 1986-era equivalent of SQLite "in" them, i.e. not actually in them because http://switchfromshapefile.org/)

2026-06-11 14:10:01 UTC

- reply

big "the self-driving cars can't help but hit pedestrians so let's just put up fences to prevent pedestrians from crossing the street anywhere" energy

2026-05-26 13:30:29 UTC

Il faut rappeler le plus souvent possible aux lecteurs mais SURTOUT aux chroniqueurs du @npub1fk4…lf9n que, en bon québécois, you do not under any circumstances have to hand it to Mathieu Bock-Côté https://www.ledevoir.com/opinion/chroniques/982007/mathieu-bock-cote-exalte

2026-03-19 14:20:46 UTC

Malheureusement on n'a plus les péquistes qu'on avait dans son temps 😞

2026-02-20 15:10:46 UTC

- reply

indeed, I guess the question is whether the scale of the *ahem* waste, fraud and abuse *ahem* of resources that LLMs seem to imply, even in benign use cases like yours, is out of line with historical precedent or not.
Am I an old man yelling at a cloud?
No, it's the children who are wrong!

2026-02-20 15:05:11 UTC

- reply

As always, yes and no. A bug zapper is designed to zap bugs, it is a simple mechanism that does that one thing, and does it well. An LLM is designed to read text and generate more text.
That we have decided that the best way to do NLP is to use massively overparameterized word predictors that we have trained using RL to respond to prompts, rather than just, like, doing NLP, is just crazy from an engineering standpoint.
Rube Goldberg is spinning in his grave!

2026-02-20 14:18:58 UTC

- reply

But ALSO: using a multi-billion-parameter synthetic text extruding machine to find spelling and syntax errors is a blatant example of "doing everything the least efficient way possible" and that's why we are living on an overheating planet buried under toxic e-waste.
If I think about it harder I could probably come up with a more clever metaphor than killing a mosquito with a flamethrower, but you get the idea.

2026-02-20 13:49:01 UTC

- reply

I agree in principle with Cory, but I really wish that he had clarified that:
1. Ollama is not an LLM, it's a server for various models, of varying degrees of openness.
2. Open weights is not open source, the model is still a black box. We should support projects like OLMO, which are completely open, down to the training data set and checkpoints.
3. It's quite difficult to "seize that technology" without using Someone Else's Computer to do so (a.k.a clown/cloud)