Table of contents

• Cooking / recipes
• Google and the web, another takeover attempt
• AI and open-source
  • “Data colonialism”
• Sad snake (Python packaging & dependency management)
• Webdev
  • Tooltips and popovers - Floating-UI
  • Observable Plot
  • Youtube to webpage
• Music!
  • Noisia - Machine Gun
  • Chancha Via Circuito - Ilaló (Ft. Mateo Kingman)
  • G Shellz & Ed Solo - Cheque

Cooking / recipes

A long-time “internet mutual” of mine finally published his mouth-watering recipes: https://jaypanoz.github.io/cuisine/ (FR).

You’ll find alot of healthy snacks recipes (a lot are vegeterian/vegan). Be warned, he is a fan of peanut butter :)

I love cooking/making food (I may post a longer article on the topic one day), meanwhile, here is two other websites I often go for inspirations:

• https://www.cuisine-libre.org/ (FR), a nice collaborative website operated by Romy. I have a few recipes on it too, I’ll let you find them :).
• https://grimgrains.com/site/home.html (EN) , the cooking website/wiki of Rek Bell and Devine Lu Linvega

Google and the web, another takeover attempt

After the AMP attempt in 2017/2018 and its fiasco, Googlers are trying to take over the web again, this time with a new “standard” proposal called “Web Environement Integrity”. It’s like the generalization of DRMs for the web, allowing a website to check if the browser is “safe” (i.e, compliant with the interests of the website editor and the browser editor… in short the advertisment/user-tracking company and private-data seller).

Mozilla/Firefox opposes this proposal and won’t implement it.

Vivaldi browser also opposes this proposal.

The web reacted quite strongly to this, and in its usual way, the issues page on the github repo is bonkers.

It should also be noted that Apple already successfully implemented a similar feature before.
Attestation mechanisms are harmful for the web and its users: this article explains the situation quite well and is worth a read — https://httptoolkit.com/blog/apple-private-access-tokens-attestation/

AI and open-source

An AI model can said “open-source”, but without its training dataset published, it’s a lie. It’s still potentially theft.

Imputability, Accountability and Transparency are key (also like in climate change management / societal ecology).

On the Accountability and Transparency topic, my friend Severo pointed me to the Hugging Face’s Open Rail licences.

On the Imputability topic, I would be quite curious to see the current state of the research. It’s a quite difficult problem: e.g. to simplify, “being able to tell, for a generative model output, which parts of the training dataset was used to generate it”.

“Data colonialism”

the process by which governments, non-governmental organizations and corporations claim ownership of and privatize the data that is produced by their users and citizens.

https://harvardlawreview.org/blog/2023/06/data-colonialism-and-data-sets/

This term is interesting and is pretty accurate imho. It’s a bit like the “data is the new oil” metaphor, but with a more political and historical perspective.

Popular large language models (LLMs) like Github Copilot and GPT-X (which basically come from the same company) are trained using massive textual datasets coming from publicly browsable web content.

That soon led to dubious TOS changes from Reddit (leading to user discontent and exodus):
which is now trying to prevent the use of its content in LLMs training
and also restrict how average users can build tools on top of its content (like alternative reddit clients).

Sad snake (Python packaging & dependency management)

At work, I regularly read, code and manage python code. I have a strong love/hate relationship with this language, mostly hate recently… :->

After being a Python dev for several years, I’ve come to the conclusion: don’t use Python.

Why that ? Mostly because of the sad state of the packaging ecosystem, leading to complicated and fragile depenency management on projects.

• https://www.bitecode.dev/p/relieving-your-python-packaging-pain
• https://www.bitecode.dev/p/why-not-tell-people-to-simply-use
• https://chriswarrick.com/blog/2023/01/15/how-to-improve-python-packaging/
• https://pradyunsg.me/blog/2023/01/21/thoughts-on-python-packaging/

I may soon post a longer article on an attempt to handle this on one of our project (at work).

Webdev

Tooltips and popovers - Floating-UI

Correctly positionning tooltips and poppers elements in webapp and webpages is a suprisingly difficult problem.

PopperJS describes itself as a “tooltip & popover positioning engine”. This library is now very popular. to help you position poppers in your web applications. A popper is an element on the screen which “pops out” from the natural flow of your application. Common examples of poppers are tooltips, popovers and drop-downs.

PopperJs’s team annouced its sucessor library: Floating-UI. It’s still new, but it looks promising and more modern (more modular, tree-shakeable and platform agnostic).

At work we use popperjs via a wrapper library: svelte-popperjs.
This wrapper does a better job than the “official” svelte wrapper library, because it doesn’t need a wrapper component and completely leverages the power of the Svelte’s actions/use directive (which allows clean and powerful integration of DOM based libraries in Svelte).

However it’s still a bit cumbersome to use, hence the news about Floating-UI. Floating-UI doesn’t have a svelte wrapper yet, but it would be quite trivial to write one.

Observable Plot

Observable Plot is a really nice dataviz library that sits on top of D3, and helps to create charts and graphs in a more declarative way.

You’ll spend more time “using vision to think” and less time wrangling the machinery of programming.

It recently gained typescript support – herm, in april – (and I hope it will soon be the case of D3 itself ^^)

Its feature set is growing quite quickly.

Youtube to webpage

I learn much better from text than from videos

A little Perl script to convert Youtube videos to webpages (thumbnails + transcripts + timpstamped links to the video).

https://github.com/obra/Youtube2Webpage

I agree with the author: the video format is not the best (for me) to learn things. Text is superior in many ways.

Music!

Noisia - Machine Gun

A classic of angry drum&bass

Chancha Via Circuito - Ilaló (Ft. Mateo Kingman)

Virgen de la madrugada
Dame claridad y tranquilidad
Lucero de la mañana
Dame luz y dame calma
(x3)

Bajando por la montaña, la luna me regaló
Un lucero en la mañana, lunita del Ilaló
Alma de la oscuridad, salvia de mi vida
Alma de la oscuridad, salvia de mi vida
Ay ay ay ay

Montes se copa tre volcano madre dile a los taitas
Coturcose rosa cae rio panocoto ayaya ay luna chiva
Cero en mi corazón de aguacero mira por eso hombre el cielo
Miran al cosmos miran al todo miran

Frailejones dorados danzan en torno a una curandera, una fuerte hogera
Rompe descalzo mis pies se queman
Ayayay ayayay vuelan

Bajando por la montaña, la luna me regaló
Un lucero en la mañana, lunita del Ilaló
Alma de la oscuridad, salvia de mi vida
Alma de la oscuridad, salvia de mi vida
Ay ay ay ay

G Shellz & Ed Solo - Cheque

Oldschool jungle/dnb, dumb and catchy - A family classic in the car ^^