Antoine Roux

Not to convince you, but just a tip: copilots are really good at writing tests (if you structure the approach). Which is a good way to put your tech debt under control. AI won't fix your tech debt, but it can help.

But you can ask your AI agent to write tests for your existing legacy code base without a test suite...

But first and foremost: average number of cheek kisses in country?

It raises questions though. How do you deal with intra-annual tasks? For instance if you want to detect the date at which a cut happened? Do the embeddings allow you to infer this information?

I think from here you are supposed to train a simple model. For instance, if you want to detect deforestation, instead of using a series of images, you will train the model using the embeddings as an input.

Paris

Were the sessions recorded?

From my perspective : no vendor lock-in, it is now the de-facto standard.

Has anyone tried Ray to run their ML training and inference tasks? Any opinions?
Please share for visibility.

Actually... It is probably one of the most interesting on the list. I wish I could try it :-D
It looks like French pâté de tête.

This isn't uncommon. I participated to a training with the French counter-espionage security agency, they warned us against this kind of situation. They gave examples of real cases that happened with Chinese candidates.

Agreed! Anchovies work so well to make a sauce stand out. One of my favorite pasta sauce is Alla Puttanesca, where the anchovies are part of the reason why it is so tasty.

Thanks!

Tu as un lien vers la recette ? Je sais qu'on en trouve plein sur Internet, mais je n'ai jamais essayé.

On peut faire une bannière don't give up ajoutée = une blague aussi

Patch « Don’t give up ! » Avec un drapeau ukrainien

#Ukraine 21/02 (21) L’idéal serait de monter et déployer un vaste mouvement d’opinion #DontGiveUp qui influerait sur les perceptions des dirigeants européens, ukrainiens et américains…
Qui en est ?

Someone from Microsoft once told me they would keep maintaining Azure DevOps, but all the new stuff goes to Github.

AWS offers repos and pipelines, not issue tracking and project management as far as I know.
Last time I checked Azure DevOps, full YAML support or pipelines wasn't available yet. Some actions had to be defined in the UI, they couldn't come from a YAML file in your repo. It might have changed now.

Alberobello

Wow this is very niche.

Time in software is one of my pet peeves. It is to good kind of crazy. So I was happy to discover that, finally, ECMAScript will be getting a proper date and time API.
tc39.es/proposal-tem...

My team's bug/feature ratio over the last 6 months: 0.14. I'm living the dream...

100% agree. I would add sushi, fresh pasta (but not fresh ravioli, those are worth the effort).

Observability for small companies/early start-ups ? If you are small and don't have the complexity of a scale-up running on micro-services, but also a very small team, what are the best things to monitor, the best tools to succeed while keeping things simple?

Playing with Python 3.13 and free-threading. I find it surprising it took so long to Python to get proper threading.
There are many occasions when threading is the right tool to solve a problem.
But I finally saw a Python process using 1100% CPU 😀

80% of the time the monolith is not the evil

The rest of the time probably still not.

Ho oui.
On analyse Trump 2016, le RN, le Brexit... Mais soit personne n'a la bonne analyse, soit personne n'y comprend rien : l'histoire se répète et on dirait que tout ça reste un mystère.
Les gens continuent d'être surpris, les partis conventionnels ne savent toujours pas endiguer ces phénomènes.

I am CTO at Cybeletech, a small company developing digital twins of plants. I make SaaS products based on these scientific capabilities. I used to work at a company doing monitoring for MS Teams. I see myself primarily as a software engineer.

Are there other examples of small companies achieving exceptional results, that are well documented?
Especially companies that are profitable. WhatsApp had no business model. It is always easier to operate at a small scale if you can just burn other people's money.

The set of trade-offs that works for large companies does not apply to smaller ones.
I read about WhatsApp using Erlang, but I suspect it is not the only reason. They were working across many countries, had mobile apps for several platforms, were handling millions of users on their backend...