Anyway if anyone was looking for the latest DNG specifications, they are available at:

Adobe: helpx.adobe.com/camera-raw/d...
ISO: www.iso.org/standard/861...

Bonus point for the Adobe one is that they seem to have published or circulated two different versions without incrementing the publication date or the spec’s version number.

So some people have a DNG 1.7.1.0 PDF that says “this is not allowed”, and others have one that say “this is allowed”.

After a couple decades reading standards from W3C et al, it’s a shock when you try to read some other specification published by someone else.

Adobe: sure, try to fish out some PDF from the unsearchable entrails of our many websites.
ISO: here it is, that will cost you CHF 227 for a 100-page PDF.

Basically trying to do a babyphone/nannycam thing. Not routing through the global internet please (I'm in the ocean, latency is high).

Anyone knows a cross-platform app that can stream audio/video from one device to another on the same wifi network? Everything I find is for phones, but I need it on a laptop too.

remake The Vampire Diaries & set it in the age of true crime podcasts, you cowards!

every time i read about ai agents allegedly blogging about drama, it’s really just some human playing wizard of oz

A good read on issue management for open source projects, which matches the consensus among Godot maintainers.

(Any patronymic resemblance to the author may or may not be fortuitous.)

Yeah, I've advocated for "it's just a number" regarding open issues too: fvsch.com/stale-bots

The thing with "not everyone likes working from home," "some people work better in company" is some people don't like working in the office & work better left to their own devices but apparently that's not as important - they're expected to be companion donkeys as well as doing their actual job.

We expected Agent Smith; we got Colin Robinson instead.

Would you like to hear it Now or Maybe Later?

Why Node.js Needs a Virtual File System Node.js now has a virtual filesystem. Learn why VFS matters for testing, sandboxing, Single Executable Apps, and running generated code directly in me

We're announcing two things:

node:vfs, a Virtual File System landing in Node.js core (PR #61478, ~14,000 lines across 66 files)

@platformatic/vfs, a userland package you can use today on Node.js 22+

blog.platformatic.dev/why-nodejs-n...

TERMINAL, by Anaerobic 6 track album

Stranded @ JFK airport, so overnight I wrote and recorded an absolutely insane noise album.

Documented events include:
• salad
• the custodian named Hugo cleaning the floors
• Delta's AI helpline
• a baby

This is real and it's real weird. Maybe you'll dig it.

anaerobic.bandcamp.com/album/terminal

I'm busy full time with childcare and prepping our move abroad, and plan to look for a job or freelance contracts right after that.

I feel like trying out some AI coding is going to be a requirement in my job search, so not sure I have the luxury to have a strict "not even once" policy.

Extra background:

- I don't like all the other ethical issues with those companies. It's not just the school bombing thing.
- Deeply wary of the impact of generated code on medium and long term software health.
- Use case that interests me is asking a LLM for rough guidance through a codebase.

So, consensus among devs who use AI coding tools is that Anthropic and OpenAI's models/agents are finally good now (since Dec 2025). It's reaching a critical mass which makes me think *maybe* I should try those. But I also don't have too much cash and don't want to fund school bombings. Thoughts?

The emerging cases of AI psychosis have shown that all it takes to drive some people completely insane is to put them in contact with a sycophantic chatbot who always agrees with them.

Now, what does that tell us about billionaire CEOs surrounded by sycophantic people who always agree with them?

This reminds me of discussions around the harms of meat production. There is a subset of meat consumers in those discussions that try to reach out to the “reasonable” vegans or vegetarians and ask them for absolution for their own super-ethical meat consumption.

Your menu doesn’t need Miller’s 7±2 rule Be careful with anyone selling “magical” UX rules as universal answers

Your menu doesn’t need Miller’s 7±2 rule. Be careful with anyone selling “magical” UX rules as universal answers.
Miller’s 7±2 rule was about memory, not how many elements you can put on the screen.

stephaniewalter.design/blog/your-me...

Our Spring Sale starts today! Save 20% until March 23rd getkirby.com/buy

You know the drill:

🛒 Volume discounts save you even more
🛒 You can buy licenses in bulk and activate them later
🛒 Each license includes free major upgrades for the next 3 years after activation

Saw a French language trailer for the movie Zeh Weel Keel Yoo.

Et pour le coup du HPI, peut-être se renseigner sur "la légende noire des surdoués" et demander en quoi ça excluerait des commodités telle que le TDAH (et d'après quel manuel diagnostique ou recommandation de panel d'experts). (Faut sans doute le formuler autrement pour pas braquer les Sachants.)

C'est nul comme situation et c'est pas sûr que trouver des sources aide :/
Mais c'est la seule idée que j'ai, à part demander un deuxième avis médical à un-e spécialiste qui se repose pas sur des tests psychométriques chers et trompeurs.

Faudrait retrouver les études mentionnées dans la première vidéo, les obtenir et imprimer au moins les abstracts. Bon après si on lit pas l'anglais ou si le psychiatre lit pas l'anglais, ça aide pas. Faudrait voir du côté des experts francophones comme Nader Perroud.

There Are No Tests for Diagnosing ADHD YouTube video by Russell Barkley, PhD - Dedicated to ADHD Science+

De mémoire il y avait une vidéo sur le sujet sur la chaîne YouTube de Russel Barkley (expert sur le TDAH). Je retrouve ça:

youtu.be/nVCDflvwkE8
et
youtu.be/5voIJu8wwXw

Faut trouver les études qui montrent que les tests psychométriques qui ont la prétention de diagnostiquer le TDAH sont très peu fiables (pouvoir prédictif dans les 60% je crois), alors que les grilles d'évaluation font beaucoup mieux (dans les 80% ou plus).

And pixel-peeping at 400-800% zoom shows no artefacts with this new “high fidelity DNG” export format, with rare edge cases where there is a slight loss of details at this zoom level in very dark shadows brought up 3-4 stops. Old JPEG can't do that indeed.

Reviewers (photographers, not software devs) talk about file size reductions compared to original compressed RAWs, e.g.:

- Original RAW (out of camera, w/ lossless compression): 35MB
- DxO lossless DNG (demosaiced + denoised image): 100MB
- DxO high fidelity DNG: 20MB