Mathieu Bastian

📣 Big news for Gephi! Today we are launching Gephi Lite v1.0 (the web version of the Gephi software). 📰 Read the blog post describing this milestone: gephi.wordpress.com/2025/10/08/gep… 💻 Or try Gephi Lite v1.0 here: lite.gephi.org

Today, we are releasing Mistral Large, our latest model. Mistral Large is vastly superior to Mistral Medium, handles 32k tokens of context, and is natively fluent in English, French, Spanish, German, and Italian. We have also updated Mistral Small on our API to a model that is significantly better (and faster) than Mixtral 8x7B. Lastly, we are introducing Le Chat (chat.mistral.ai), a chat interface (currently in beta) on top of our models.

New post! Picking at some of my stream processing scar tissue. Why Samza failed, how it led to Kafka Streams and Kafka Connect, and why I'm skeptical of Apache Flink. materializedview.io/p/from-samza-t…

Built something I like, and now it’s out for anyone to try for free! ⚡ Introducing Peek - a neat way to summon ChatGPT, Bard, Perplexity anywhere on your Mac. Details 👇🏼

@OfficialLoganK @OpenAI @andrwpng Yes, would be immensely useful as we use one key per use case. It would help governance and better forecast future spend

Would you be excited / would it be useful if we released usage tracking based on API key for the @OpenAI API? 🤔 cc @andrwpng

Shipped Peek with an Intel version for those running that on Mac. Here you go: prateekkeshari.gumroad.com/l/peek

📢 New research alert! 💡 We've developed PRODIGY: pretraining framework for in-context learning over graphs. Through a novel prompt graph representation and a family of in-context pretraining objectives, our model can adapt to novel tasks on unseen graphs 📈. Outperforming contrastive pretraining baselines by 18% and standard finetuning with limited data by 33% on average, PRODIGY proves its strength in citation networks and knowledge graphs. Read on: arxiv.org/pdf/2305.12600… #AIResearch #GraphLearning #PRODIGY #InContextLearning Joint work with @qhwang3 @ren_hongyu PengChen GregorKrzmanc DanielZheng and @percyliang

We’re proud to share that we’ve secured $194m in new funding! 🚀 We’re excited to work with our team and supply partners to further accelerate our mission to unlock unforgettable experiences for travelers around the world!

Of all the announcements in the pipeline, I’m especially excited about first class support for LLM’s of all flavors in MLFlow. Hugging Face, OpenAI and Langchain all play well with Databricks. databricks.com/blog/2023/04/1…

Pandas 2.0 is here! This is the biggest overhaul of Pandas since its inception, and it has been years in the making. However, you will probably not notice too many changes, and all your existing Pandas code will most likely run the same as before. All the major changes are under the hood. That's because Pandas has moved away from the way it represents data, from numpy to Apache Arrow. Pandas was originally built on top of numpy, and it was an adequate solution for many tasks. However, three are many limitations of numpy that have only become more obvious over the years. Apache Arrow will significantly help with those pain points, and will speed up many Pandas tasks. I've only played with the new version for a day so far. My limited impression is that it significantly speeds up loading and saving of csv files, and puts the new version of Pandas on par with Polars in that regard. Lookign forward to playing more with it in the weeks and months ahead. Great blog post about what's new in Pandas: datapythonista.me/blog/pandas-20… Release notes: pandas.pydata.org/docs/dev/whats… GitHub repo: github.com/pandas-dev/pan… #DataScience #MachineLearning #Data #Python

Check out Gephi Lite!

🎉Big news!🎉 Introducing 🚀Gephi Clique - the ultimate network visualization tool! 💥Unleash the power of our game-changing SUBSCRIPTION MODEL! 🤯Join the Clique today and revolutionize the way you visualize networks! #DataAnalysis #JoinTheClique 👇👇👇 gephi.org/clique/

Excited to share our new paper, ‘Accelerating network layouts using graph neural networks’ just out in @NatureComms with @csabath95, @nimatabari, and @yuqirose: rdcu.be/c77wd NeuLay is a machine-learning framework to speed up network visualization. See thread:

Maybe Instruction is All You Need? It seems to have an outsized impact vs scale and even some aspects of data selection and curation to get a usable instruction-following model. We are excited about what this means for democratizing LLMs and letting every org build its own.

Gephi 0.10 is out! • Quick search • Dark mode • Apple Silicon support • Improved image export • Workspace management • More bugs fixed Check out the release post: gephi.wordpress.com/2023/01/09/gep…

@felipevallerio @criccomini No, not sure if its still used or if someone took it over.

I want embedded data stored for everything. SQLite and DuckDB have spoiled me. I want this for all types of data stores: JSON document stores, graph DBs, search indexes, embedded Redis, etc etc. (JVM doesn’t cut it. Needs to be portable like SQLite and DuckDB.)

@seinecle @nipafx @gunnarmorling Not an easy fix but you could define your model as OpenAPI and have it generated automatically at build time

cc @mathieubastian @nipafx @gunnarmorling

#Java question: I have two projects A and B: an API endpoint, and a webapp that makes calls to it. Both projects have a dep to a 3rd project, which is the model of the data sent / requested to the API. Problem:

The @Gephi paper has just reached 10.000 citations, congratulations to its authors @jacomyma @mathieubastian @mentatseb and thanks to those who continue to develop this essential software today! (scholar.google.com/citations?view…)

Back in August as part of the @Gephi week I recorded an intro video to its architecture and codebase youtu.be/0OVQQACjLck