HashiCorp co-founder and Ghostty creator Mitchell Hashimoto, now a co-founder of Superlogical, takes it even further. He thinks “the future is that large open source projects will close contributions completely.”

Ruiz responded, “It just makes less sense to have people contributing code if the issue is decently well-specified and the code can be written by agents.”

and this is how the open source could die by closing itself and "the code can be written by agents".

Developing issues instead of MRs may be a solution.

TL;DR: we spend more CPU cycles rendering commits for scrapers than we spend on all other kinds of legitimate access, including git clones. At any one time, across 5 geo-distributed nodes, there are 14 CPU cores doing nothing but rendering git commits as html.

AI crawlers are idiot: their are consuming the HTML diff of the commits insteadof cloning the repository locally...

Building websites by hands mean the solutions to errors are within you somewhere. "You’re also rewarded with greater understanding of the problem and solution, knowledge that you’ll carry with you."

An AI agent hides, obscures and obfuscate the result or the method used to reach such result.

The web can florish with little websites everywhere. It's never about the size of the audience for the OG web, but always about the community.

Lol. The second parody of cats.txt made by another person.

It took off and https://catstxt.org/ was born.

On the contrary, llms.txt is often ignored by the crawlers.

The trap is this: getting baited into treating a set of observations as evidence, when the observations would occur whether or not the underlying thing were true. It is the intellectual equivalent of concluding your umbrella causes the rain to stop, because every time you put it away the rain does eventually stop.

It's a convenient example: . The same broken chain of inference is being applied to almost every new GEO tactic invented on a monthly basis

When you ask a model whether llms.txt helps, it is not reasoning. It is not running an experiment, consulting a source or weighing evidence. It is returning the most common thing it has seen written on the subject. The web is thick with confident posts declaring llms.txt the future, so the model converges on that consensus and reflects it back, dressed as a considered opinion. It endorsed my cats for precisely the same reason: by the time anyone asked, enough people had written enthusiastically about cats.txt that the average of the discourse said “yes.”

Quite the opposite of GEO: " That is what the “O” in SEO is meant to stand for. Optimisation is the cumulative advantage of doing the small, verifiable things a little better than your competitors, over and over, until it adds up."

Quelques problèmes des affiches générées par IA:

• toutes les informations sont mise au même niveau
• toutes les affiches s'uniformisent. Comment se démarquer des autres ?

Ce style explose, tout comme les affiches de type PowerPoint ou Word s'étaient démocratisée auparavant.

Github stats are not pertinent anymore because Claude AI can turn them upside down.

It’s now easy to ask Claude to generate a list of interesting open source projects, then ask Claude to find some problems in them, and then ask Claude to raise some PRs to fix them. You don’t even have to use the projects or care about them, but you can easily create the illusion to outsiders that you care, or that you found a problem, or that you put the time into fixing it. [...]

Maybe this was unreasonable, but truthfully, I’m just not interested in encouraging people to take up our time with this kind of busywork. I do not want to set a precedent of accepting PRs that materially improve nothing, nor do I want our contributor list to become a reward for asking a robot to fix typos.

The bet is AI can rewrite large portion of codebases. Let's see.

How AI is impacting negatively the OSS maintainers. It's written as a story.

Demander à l'IA de saisir les tâches dans un fichier markdown afin de pas les ignorer.

Si tu découvres un problème ou une erreur pendant une tâche, même préexistant(e) et hors du périmètre de la tâche en cours, ne l’ignore pas silencieusement : soit tu le corriges, soit tu ajoutes une description détaillée dans un fichier TODO.md à la racine du projet, pour ne pas perdre cette découverte.

Some thoughts about the AI watermarking:

1. the way it's implemented makes it hardly possible to verify the watermark independently of the LLM's owner.
2. It helps the monster to differentiate fresh food from its own vomit, so it will not feed from the low-quality slop it once created.
3. Something that is not immediately visible as a human, and easily removed automatically is basically the opposite of a watermark.

Une personne meure aléatoirement sur terre pour 10 000€ est un dilemme moral, mais appuyer sur un bouton et consommer plein d'eau, d'énergie, contribuer à faire monter le fascisme et les températures, en échange d'une image ou d'une mélodie plagiée; c'est OK.

The agent publish the news of the website daily.

Alors qu'en janvier 2026, Elon Musk twittait « Grokipedia connaît une croissance fulgurante. Elle dépassera largement Wikipédia cette année »