
Tech lead. Wannabe Design Engineer. Organizer @AperoWebNancy
Joined June 2011
- Tweets2.2K
- Following472
- Followers115
- Likes339
New blog post: Self-hosting LLM models for software development
It's surprisingly easy to self host a medium-sized model.
Cheaper than I expected!
Performances are good enough for me to reduce my Claude personal subscription.
kmaschta.me/blog/2026/09/22/…
Valentin Manceaux-Panot retweeted
Defining row-level security policies can be simple and efficient!
✨ acadia.engineering/blog/simp… ✨
Acadia even does "symbolic evaluation" at compile time so that in many cases, you pay zero runtime cost for your security guarantees!
The example app runs an Elm server too! 🚄
Valentin Manceaux-Panot retweeted
Acadia gives some interesting guarantees:
- No 1+N queries.
- All queries terminate.
Learn how a query language from the 1970s (Datalog) inspired the key design choices that make these guarantees possible in Acadia's query language! acadia.engineering/blog/solv…
Valentin Manceaux-Panot retweeted
Acadia is out today! Bringing the benefits of @elmlang to SQLite and PostgreSQL:
- Friendly error messages
- Custom types for precise column data
- Simple module system for easy reuse
Check it out! acadia.engineering/blog/reth…
Valentin Manceaux-Panot retweeted
I made a site for artists to get museum-level references since Pinterest has been a bit frustrating lately.
Open forever.
Free forever.
No AI images ever.
scry.studio/
Valentin Manceaux-Panot retweeted
Thank you @markzhangart ! I'm a 3D artist with an interest for lights and environments
Valentin Manceaux-Panot retweeted
Tragically I am continuing to find that the most effective guardrail against slop is extremely talented engineers doing very thoughtful, human code review
Valentin Manceaux-Panot retweeted
I guess this isn't clear: that's what code is. you're writing instructions that describe how software will look and work. that's what we're still doing in this brave new world.
Valentin Manceaux-Panot retweeted
I wrote about a mundane agent PR I almost merged last week. it fixed the bug, passed review, and made the codebase a little bit worse. there was nothing remarkable about it...and that's kind of the point. danphilibin.com/how-a-codeba…
Valentin Manceaux-Panot retweeted
L'Apéro Web de Mars aura lieu mercredi 25 à 19h! On verra comment maximiser l'efficacité des agents de codage
meetup.com/aperos-web-nancy/…
Valentin Manceaux-Panot retweeted
L'Apéro Web de février ça sera le mercredi 25 ! On se retrouve à la Plantation pour voir comment encaisser 10000 requêtes sur une VM raisonnable
meetup.com/aperos-web-nancy/…
Valentin Manceaux-Panot retweeted
The reason why RAM has become four times more expensive is that a huge amount of RAM that has not yet been produced was purchased with non-existent money to be installed in GPUs that also have not yet been produced, in order to place them in data centers that have not yet been built, powered by infrastructure that may never appear, to satisfy demand that does not actually exist and to obtain profit that is mathematically impossible.
Valentin Manceaux-Panot retweeted
📣 Pour l'Apéro Web de décembre, pas de speaker parce que c'est Noël ! On se retrouve quand même le mercredi 17 (attention, ce n'est pas le dernier du mois), à 19h !
meetup.com/aperos-web-nancy/…
Valentin Manceaux-Panot retweeted
with background agents you too can turn a 2 minute fix into a 30 minute task
Valentin Manceaux-Panot retweeted
📣 L'Apéro Web de Novembre aura lieu le mercredi 26 Novembre à 19h. On se retrouvera de nouveau à la Plantation pour parler de pull request!
meetup.com/aperos-web-nancy/…
Valentin Manceaux-Panot retweeted
hahahaha the bed sends 16gb of data a month oh god
Replying to @m_franceschetti
Is 16+gb/mo a normal amount of telemetry? Can you not do any local compute of “get hot” or “get cold” with a multi core processor and multiple gigabytes of memory? Can’t just repeat the previous nights settings?
It’s bad enough that you slapped a $200/yr subscription on things, worse that it doesn’t work at all without internet.
Valentin Manceaux-Panot retweeted
hey look, an AI artist got hurt because somebody stole his prompt
Never do this:
Passing off someone else’s work as your own.
This Grok Imagine effect with the day-to-night transition was created by me — and I’m pretty sure that person knows it.
To make things worse, their copy has more impressions than my original post.
Not cool 👎
Readers added context they thought people might want to know
"Day-to-night transition" is a basic AI prompt for time-lapses, used in Runway AI (Oct 2024) & Stable Diffusion since 2023. Not proprietary—sharing isn't theft.
tomsguide.com/ai/ai-image-vi…
Valentin Manceaux-Panot retweeted
Maybe it’s my algo, but in all this talk lately about how critical it is to hire amazing designers (it is), I’m not seeing many folks mention the other half of the requirements - amazing engineers.
Probably every one of those top tier designers you’re wanting to work with have experienced working with an engineer who doesn’t work at the same level as expectations and ambition. This is a near insurmountable burden to a non-technical designer, because they’re limited to the capabilities and ambition of their engineers. My toughest career experiences have always involved being limited by an engineer’s capabilities rather than a designer’s.
While yes, top designers will seek cultures where quality and craft is a top priority, but they'll need the same expectation on the engineering quality to support their ambitions. Otherwise, the quality bar only lives in Figma.
For example:
- Optimistic updates
- Ultra fast queries
- Caching optimization
- Component api DX and maintainability
- Concurrency and parallelism
- Network optimization
- Build tooling and pipelines optimizations
- Low (zero) dev env friction tolerance
None of these things involve design, yet they’re the foundation of a design’s implemented quality level. Other than what a user see’s, these are the parts of the design a user actually experiences. These types of things are the secret sauce that set good applications from great ones. These are table stakes for great engineers, and the top tier designers you’re looking for are also looking for teams with these engineers. Great engineers compliment great designers. Great designers compliment great engineers. Together, they turn imagination into innovation.
Designers create the “What if?”. Engineers control the “Why not?”.





























