Loïc Delmaire
@skelz0r
Followers
746
Following
6K
Media
105
Statuses
3K
Co-founder @hellojam_ & @SuperdocuHQ Currently working on API Entreprise/Particulier/DataPass @_DINUM & https://t.co/52pOLiVWkz ✏️ https://t.co/9FhqsJltuP
Bordeaux, France
Joined February 2012
Even if current LLM progress hits a brick wall at Opus 4.5 level (and I doubt that will happen) the next 12 months are still going to be a staggering time of change in this industry as decision makers start truly understanding the new reality we live in. https://t.co/cB2iOcHcN6
obie.medium.com
Something that’s been true for me for a long time, long before coding agents showed up, is that the initial effort involved in any…
17
45
570
I migrated cursor.com from a CMS to raw code and Markdown. I had estimated it would take a few weeks, but was able to finish the migration in three days with $260 in tokens and hundreds of agents. Here's how I did it + all my my usage stats. https://t.co/QIAOmLsffx
leerob.com
$260 in tokens and hundreds of coding agents later.
263
248
4K
Vibe Coders: fyi, when you fix a bug, prompt the AI to make a "regression test". This captures the bug and should prevent the AI from making that exact same mistake in the future. They are *extremely important* and I highly recommend prompting for them whenever you fix anything.
107
154
3K
2026 could be the last year where I'm going to write code for living.
2
0
1
Most of coding was never about writing code. AI is just making this more obvious. You no longer need to recall syntax, function structure, boilerplate code, or even API endpoints. That’s the easy part and AI is very good at it. The hard part was never typing. It was always
320
434
4K
Put together some notes on Mistral Vibe, Mistral's new CLI coding agent (think Claude Code / OpenAI Codex) which is Apache 2 licensed, written in Python+Pydantic+Textual and has a neat set of system and tool definition prompts that are fun to read
simonwillison.net
Here's the Apache 2.0 licensed source code for Mistral's new "Vibe" CLI coding agent, released today alongside Devstral 2. It's a neat implementation of the now standard terminal coding agent …
11
44
407
Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵
144
451
3K
Le président Emmanuel Macron était en déplacement officiel en #Chine cette semaine. Un voyage dont on parle beaucoup, du fait de l'actualité et des enjeux auxquels l'Europe fait face dans le nouveau jeu économique mondial. Un voyage que j'ai moi-même fait il y a peu au cours du
81
312
856
If a $20k MRR app can really sell for ~$1.4M… Then the playbook becomes obvious: Build 4–5 products → exit → put everything into the S&P 500. Monthly returns, zero stress, no meetings, no deadlines. Just ships → exits → dividends. Honestly… this might be the simplest
101
37
1K
Raw css should be more commun
Today's Fizzy deep dive is CSS "While the industry sprints toward increasingly complex toolchains, 37signals is walking calmly in the other direction." https://t.co/lNsAkEzbxC
0
0
1
Unlike Campfire, where we shared a zero-history copy of the codebase, Fizzy's public codebase has the entire life of the product development cycle out in the open. 1,792 pull requests of historical fun and insight into how we make things at 37signals!
26
44
1K
I'M LOOKING FOR A TECHNICAL PARTNER I'm looking for a coding wizard to help me build software (and to bounce ideas with). The deal is simple: you build / I distribute You do all technical work. I do all marketing, distribution, customer research. What I bring to the table:
174
42
464
INTRODUCING FIZZY Have you noticed that every issue and idea tracking tool you loved slowly morphed into boring, sluggish, corporate bloatware? Trello put on 40 pounds of cruft. Jira started charging by the migraine. Asana tried to become everything to everyone. GitHub Issues
191
134
2K
We put together a prompting guide for Claude Opus 4.5 based on extensive internal testing by our research and applied AI teams. Here's what we've learned so far about getting the best results:
63
228
3K
My experience with Gemini 3 Pro was mostly negative. Too much thinking, too little action, went on side quests very often with overconfidence. Back to Opus 4.5 with Amp's `--try-opus` and the first two sessions were perfect. Example bad (Gemini):
ampcode.com
40
25
626
It's been a friction point from the start but I'm really starting to think that attention management is one of the big challenges of agentic programming. And I'm talking about human attention.
29
2
147