Jordi Cabot
jcabot.bsky.social
Jordi Cabot
@jcabot.bsky.social
190 followers 44 following 190 posts
Head of the Software Engineering RDI team at LIST (Luxembourg). FNR Pearl. Former ICREA, UOC, Inria/IMT, U of Toronto. 2xFounder. Interested in how to build better software faster using a combination of low-code, AI and open source
Posts Media Videos Starter Packs
☄️First full draft of my new #book 📖 : "How to survive (and thrive) in research". With 150 #advices for #PhD and #junior #researchers

Now I'm going to go over all your comments and suggestions 🔎🕵️ and share the draft with early readers

Learn more ➡️➡️ jordicabot.com/book-advice-...
🕸️Visit our new brand website🕸️

🔗 besser-pearl.org

Many things to improve, but it's a start!

#lowcode #opensource #modeling
🥰 Always happy to get this feedback 🥰:

"We are adopting BESSER. We had the 1stclass, and it was easy for students to start modeling (much more than with EMF"

Already 20 univ. using BESSER to teach #mde #uml #modeling

Do you want to be next? ➡️ lowcode-book.com/book-users/
Forget #vibecoding , embrace #vibemodeling.

And soon this will be possible in BESSER. See the #teaser below.

For more on vibe modeling --> modeling-languages.com/vibe-modeling/
This is what we tested in a recent work where we show that current LLMs beat fine tuned versions of older ones in the education domain, in particular in question generation tasks.

And we've seen this also in other domains where top finetuned models on a task get obsolete by newer general #LLMs.
Should you #finetune or should you wait?

Finetuning is still an option to improve the quality of the results, but a costly one (in terms of data, time and expertise).

So if you can #wait, do it. It is likely the next generation of #LLMs will be good enough for you.
A 2024 paper with already 10 citations 🥳🎉

Glad that you do find the topic of #LLMs for #modeling interesting!

Haven't read it yet? See how we use a #ToT strategy for #domain modeling scholar.google.es/citations?vi...

#uml #llm
🚀 Exciting new feature in release 4.2.0!

We just integrated the #IMG2UML feature into the graphical editor of BESSER 🎉

This allows users to upload an image of a class diagram and automatically convert it into an interactive model.
So I had to manually enter my #lowcode book into #GoogleScholar as it seems Google doesn't like self-published books and does not automatically index them... scholar.google.com/citations?vi...
scholar.google.com
ModelFed: a Federated Protocol for Collaborative Cross-Platform Modeling modeling-languages.com/modelfed-fed...

inspired by the #fediverse and #actvitiypub
📢 NEW #BOOK PROJECT📢

Read about my new book project aimed at giving #junior #researchers the advice they need to #accelerate their #research #career 🏎️

🔎 jordicabot.com/book-advice-...

Answer the short survey📋in the post to make sure the book is useful to you.

Pls RT

#phdchat #phdlife #postdoc
Glad to see that @OutSystems and I agree on the importance of vibe modeling modeling-languages.com?s=vibe+model... (vs #vibecoding).

See this interview to their new CEO diginomica.com/outsystems-n...
vibe modeling - Modeling Languages
modeling-languages.com
Absolutely. We are losing our minds. And conferences (not only workshops) with more and more pages to fill also generates additional stress and work as people feel bad when not filling the space (and reviewers use it to say "you could have done more eval, you have space")
Now it turns out that even workshop papers need to come with empirical validation. Next time an idea tries to pop up in my brain I'll stop it and tell her you need to come up validated or stay out of my head 😫😩