But the real pattern is more disturbing. Our research found:

• AI-generated code contains 322% more security vulnerabilities
• 45% of all AI-generated code has exploitable flaws
• Junior developers using AI cause damage 4x faster than without it
• 70% of hiring managers trust AI output more than junior developer code

The physics of software collapse:

• Modern software is built on towers of abstractions, each one making development "easier" while adding overhead. Each layer adds only 20-30% but it compounds a handful and the overhead becomes 2-6x for the same behavior.
• Energy is not infinite
• More money goes to infrastructure to support the grow. GAFAM are spending 30% of revenue on infrastructure (historically 12.5%).
• Senior developers won't exists if one can never be junior for a start

Shifting the priorities:

• when is a calculator leaking 32GB normal?
• why do we trust AI-generated code more than junior developers?
• how many abstraction layers are actually necessary?
• what happens when we can't buy our way out anymore?

The author describes a path forward and describe each points:

• Accept that quality matters more than velocity.
• Measure actual resource usage, not features shipped.
• Make efficiency a promotion criterion.
• Stop hiding behind abstractions.
• Tech fundamental engineering principales again.

Le sentiment d'injustice est encore alimenté par le fait que la part de l'héritage dans le patrimoine ne cesse d'augmenter : aujourd'hui cette part est de 60% alors que dans les années 1970, c'était 35%.

L'héritage reste pour les Français chargé symboliquement. C'est rarement juste des euros ou une maison, l'héritage renvoie souvent au travail de toute une vie, à des sacrifices des générations précédentes.

L'idée serait donc de taxer les plus gros héritages, notamment ceux qui parviennent à contourner l'impôt, via des systèmes d'optimisation fiscale, de donations, d'assurances vie ou des holdings patrimoniales. Reste à savoir où mettre le curseur car, aujourd'hui, la plupart des héritages - en tout cas pour les transmissions directes - bénéficient d'un abattement de 100 000 euros. De fait, 85% des héritages ne sont pas imposés.

Note: "The key insight is that loading massive files into memory all at once is rarely a good idea, even when technically possible."

1. Read the whole file and call it a day with fs.readFile
2. Read by byte chunks with file handles and buffers
3. (preferred) Use stream

The streams handle the bytes and UTF-8 encoding automatically. Each state can be defined with a callback: error, data, end.
Use appropriate buffer sizes for streams. Default is 64kB. The highWaterMark can adjust it though.

When to use streams: large files, real time data processing, memory is limited, composability, unknown file sizes.
When to use file handles: precise control.
When to use fs/promises: known small files. readFileSync() and writeFileSync() are acceptable for cli tools and scripts, but forbidden for web servers and concurrent apps.

Prefer Promise.all or Promise.allSettled to leverage concurrency.

Handle specific errors. Always. await handle.close() or stream.on('error', () => stream.destroy()). Another solution is await pipeline(source, transform, destinationl)

A nice python project.

It marks the function as never returning.

In comparison, the unit type () returns at least a value.

One way sighted people can determine what to put in alt text: Imagine you sent the post or article to a friend, and the image didn't load. What would someone need to know to get the equivalent experience and context? This will help you identify the important details.

TL;DR; "Humans aren’t very efficient movers—until you put us on a bicycle, when we become some of the most energy-efficient land travelers in the animal kingdom."

The content is similar to The HTML Output Element from Matthias Ott or HTML’s Best Kept Secret: The

Tag — Den Odell

.

They advocate for better usage of <output> instead of aria attributes. Let's use native instead of ARIA!

Under accessibility, but this tag should be the norm

300 billets par jour pendant 289 jours, c'est le minimum pour être rentable face à des géants en 2025.

Afin que 1.2. Train devienne rentable.

Le développeur évite de pousser à l'échelle, d'obtenir des investissements, mais de continuer le développement.

Il n'y a jamais eu autant besoin d'un site spécialiste qui ne fait qu'une seule chose. Du billet de train et uniquement du billet de train.

Pourquoi donner des milliards d'aides d'État aux entreprises privées ?

Alors qu'on pourrait simplement les applaudir à 20h tous les soirs !

Cela a tellement bien marché pour le personnel soignant.

Ok, the website has an editor built-in to change the source code. Why? Because @TheRealPomax@mastodon.social can.

(via https://mastodon.social/@TheRealPomax/115387167941878142)

Starlight has the best core web vital in A11y and SEO compared to MkDocs, Sphinx, VitePress, Nextra, GitBook, Docusaurus, Docsify.

"Speak english", "Speak understable English" and "Bonjour" on a European map