Blog
All posts including articles and notes. (2509 total)
-
Permalink
To be honest, I spent waaaay too much time in the last 2 months, probably a few hours every day, I have been on a tunnel of incremental changes and as I see things taking shape it gets really exciting :)
-
Finally some colors on my blog, I’m getting there : a place where I like to spend time, writing, reading, exploring with just enough features so that it covers my needs.
-
Permalink
Currently exploring confab/confab-web to extend the system so that we can extract learning from code sessions and push it to a documentation system
-
AI: T1Permalink
Jeremy Keith writes :
But you lose the learning. The idea of a cybernetic system like, say, agile development, is that you try something, learn from it, and adjust accordingly. You remember what worked. You remember what didn’t. That’s learning.
Outsourcing execution to machines makes a lot of sense.
I’m not so sure it makes sense to outsource learning.
I think this is already becoming a real issue in many workplaces experimenting with agentic development.
When agents generate or modify code, we can test the output. We can verify that the bug is fixed or that the feature works. But the learning loop becomes blurry: how do humans actually understand why the fix works and how the system reached that solution?
In several cases on our side we ended up doing entire review sessions where we read through the agent session logs and the final code diff after deployment. Not just to confirm the bug was fixed, but to reconstruct the reasoning behind the fix.
That reconstruction step matters because it’s where human learning normally happens in software development.
If execution becomes automated but understanding disappears, we risk creating systems that work but that fewer and fewer people actually understand.
There is probably a real need for new tooling here: something that treats agentic development sessions as first-class artifacts. Not just code diffs, but structured traces explaining decisions, iterations, and why certain approaches were abandoned.
In other words: if AI handles more of the execution, we need better ways to preserve the feedback loop for humans.
-
YGGtorrent — Fin de partie — YGGLeak
https://yggleak.top/fr/home/ygg-dossierPermalinkPutain ça rigole pas ici…
je savais pas que c’était aussi rentable comme truc … 😯
-
2 months … I went from micro.blog to self hosted indiekit and having my own ActivityPub instance via Fedify Its quite a change that now everything I use in terms of UI/backend can be changed, fine-tuned, adapted at will.
Now I’m wondering, where does it end?
well…it ends when the plugins I forked and adapted to my needs are individually reviewed to make sure they abide to upstream coding practice, but also reviewed for code duplication, code security and overall coding best practices.
Some plugins are easy to handle, others like ActivityPub/Fedify implementation or Microsub are much larger code base that will require careful review.
And then there is the UI for ActivityPub and Microsub, in both cases I followed existing indiekit approach but I still had to make bold choices that I’m not 100% happy with.
And then there is the question of contributing back to upstream knowing all the plugins and sometimes improvements or bugfix I handled where assisted and made possible in such short timewm, by Claude Code.
-
Well… I have serious doubts about this
- US decided to attack because Israel was already doing it and the US didn’t want to suffer a “blow” without also blowing things on its own
- Trump regime is stuck with a “partner” that is as rogue as the Trump regime itself, both are throwing things on the wall to make them stick, for different reasons, but in common they share one thing : survival, political survival.
Trump is running away from Epstein and Bibi is running away from himself and his war regime, they both need war to exist, they don’t even need a goal or an excuse, they just need war.
Reality is Trump is playing catchup with Israel and what have started a few days ago will last for generations to come.
-
https://blog.joinmastodon.org/2026/03/a-new-share-button/Permalink
Implemented on my Eleventy/Indiekit theme !
-
Every post on this site carries structured AI transparency metadata — visible both to readers (as a disclosure badge) and to machines (as Schema.org JSON-LD). Here’s how I built it, from the post editor to the structured data output. My AI Transparen...
Read more → AI: T1/C1 -
AI: T2
Just added AI transparency metadata to my Micropub MCP client. Every post can now carry disclosure fields for how AI was involved in creating it:
ai_text_level: 0 (None) → 3 (AI-generated, human reviewed)ai_code_level: 0 (Human-written) → 2 (Primarily AI-generated)ai_tools: which tools were used (e.g. Claude, Copilot)ai_description: free-text usage note
These fields flow through Micropub as mf2 properties and are stored alongside the post. The idea is simple: be transparent about AI involvement in your content, at the post level.
micropub_create (MCP)(type: "note", content: "Just added AI transparency metadata to my [Micropub MCP client](https://github.com/rmdes/micropub-mcp). Every post can now carry disclosure fields for how AI was involved in creating it:\n\n- `ai_text_level`: 0 (None) → 3 (AI-generated, human reviewed)\n- `ai_code_level`: 0 (Human-written) → 2 (Primarily AI-generated)\n-`ai_tools`: which tools were used (e.g. Claude, Copilot)\n- `ai_description`: free-text usage note\n\nThese fields flow through Micropub as mf2 properties and are stored alongside the post. The idea is simple: be transparent about AI involvement in your content, at the post level.", category: ["indieweb","micropub","AI","coding"], syndicate_to: ["https://bsky.app/profile/rmendes.net","https://rmendes.net/"], ai_text_level: "2", ai_tools: "Claude", ai_description: "Co-drafted with Claude Code via Micropub MCP client") Post created (202)! Posted: https://rmendes.net/notes/2026/03/03/8a851 — syndicated to Bluesky and ActivityPub, with AI metadata set (ai_text_level: 2, ai_tools: Claude). -
I now display AI usage (or not) on every post I create on my blog.
I forked
indiekit-endpoint-postsand added metadata properties that I can fill when creating a post. These properties generate their own Frontmatter and display:- An AI badge visible even before clicking into the full post view
- An AI Usage box below the post specifying whether the text or the code was produced with AI assistance or not
This is aligned with my AI transparency page.
Current AI Usage Breakdown
Level Label Posts % 0 None (human-written) 1,992 99.3% 1 Editorial assistance 11 0.5% 2 Co-drafted 2 0.1% 3 AI-generated 0 0.0% — Unset 0 0.0%
99.3% human-written. 13 posts used AI at some level (editorial assistance or co-drafted), and zero were fully AI-generated.
-
https://status.claude.com/incidents/s0pmy4yywshvPermalink
API Error 500, its Friday for half the planet already !
-
Released v2.4.0 of my ActivityPub plugin — unified the reader and explore processing pipelines into a single shared path. Posts now go through the same enrichment (avatars, media, quotes, hashtags) regardless of whether they come from your home timeline or the explore feed. Also fixed a race condition on the muted users index that could cause startup errors.
-
Just shipped two new features for my GitHub page: a searchable starred repos browser that syncs all 5k+ stars via GraphQL and caches them in MongoDB, with filtering by language, topic, and full-text search — and GitHub Lists support with tabs, sorting, and filtering. Finally easy to find that one repo you starred months ago.
mas.to
adactio.com
jacobin.com
newrepublic.com
status.claude.com
birchtree.me