Quentin de Quelen

Unveiling our new startup Advanced Machine Intelligence (AMI Labs). We just completed our seed round: $1.03B / 890M€, one the largest seeds ever, probably the largest for a European company. We're hiring! [the background image is the Veil Nebula - a picture I took from my backyard, most appropriate for an unveiling] More details here: techcrunch.com/2026/03/09/yan…

@kingofknowwhere @unk_data Meilisearch can fix this with its hybrid search.

Yeah that's because they use fuzzy matching with tolerance (check polyfuzz for ex). The problem is that this is a uniquely Indian problem of extreme misspelling for ex: red bull spelt as redbul or redull would be discoverable but it's hindi phonetic pronounciation ranges from raid-bool to rayd-bewl. That coupled with English typing (this is why Zomato has pictures, finite search options and partial word searches). This is something any fuzzy matcher will not reliably predict. An LM style solution reliably fixes this.

If you're an engineer at zepto / blinkit you know how big multi lingual search is for ex: people know the names of vegetables / fruits in their own language. This fixes it and you don't have to do a multilingual meta tags marking for new products in catalogue. Absolute genius

@liveblocks Congrats on the move!

We've open sourced our sync engine and dev server. Starting today, you can run and test your multiplayer applications locally using our new server package and CLI. Additionally, this marks a shift towards an open-source Liveblocks. lblcks.io/nys2qqB

Once again, Rust is the solution to the crap that happened to MongoDB. People who cannot understand the reason should take a step back: #mongobleed is no more than #heartbleed, and is due to the use of an unsafe memory language. Use Rust, not C, C++, or other crappy languages.

Yes! Rust is better than C++ and one of the reasons is because it landed in the Linux Kernel. C++ tried and failed... lwn.net/Articles/10498…

Introducing the Devstral 2 coding model family. Two sizes, both open source. Also, meet Mistral Vibe, a native CLI, enabling end-to-end automation. 🧵

🚀 Our EuroRust 2025 talk is live! Our CTO, @Kerollmops, breaks down how we build Meilisearch – a lightning-fast, open-source search engine powered by Rust. 🦀 Real infra lessons ⚡️ Low-latency architecture 🔍 How Rust enables our speed and safety 🎥 Watch: youtube.com/watch?v=ULA1S-…

A new video is on YouTube now 🙌 Watch @Kerollmops's talk about how @meilisearch has been using Rust to power fast open-source search! 🦀 See the full talk 👉 youtube.com/watch?v=ULA1S-… #RustLang #RustConference #EuroRust

Why not use the hybrid search of Meilisearch? I'm not sure which version we are using, but we are releasing numerous new features weekly. Additionally, you'll benefit from auto-embedding and caching. I also highly recommend using the new experimental vector store, which is much more scalable. #vector-store" target="_blank" rel="nofollow noopener">meilisearch.com/docs/reference…

My suggestion is to use a 100 MB payload and not 1000 record updates. Use Binary Quantization because it will be much faster, and you don't need that much precision if you keep hybrid search powered by Meilisearch. Then you can look at #get-tasks" target="_blank" rel="nofollow noopener">meilisearch.com/docs/reference… and the field progressTrace and internalDatabaseSizes to understand what might take time. Claude can do that :)

@Quentin_dQ @abhi1thakur This is what I get when I try to index and process meilisearch on 10% of my data:

Let's make search great again 🚀 I'm joining Vespa ai and my role is to help you build a world class search, RAG and recommendation systems. In addition, I will be: 👨🏽‍💻 Taking part in open-source communities and contributions 👀 Creating tutorials to learn how to build search for real world problems 📈 Working with cross-functional teams to bring search from "good" to "great" for real users 👨🏽‍🎓 Learning and evolving: as always If you’re passionate about search, rag, embeddings, hybrid architectures, or you’re using (or exploring) Vespa ai and want to collaborate or chat, please reach out! Always happy to connect.

@kennyhyder @abhi1thakur What do you mean by didn't work? The vector storage is now GA, but we have a new experimental one that is 100x more performant. You activate it on the index settings #vector-store" target="_blank" rel="nofollow noopener">meilisearch.com/docs/reference…. This one is really a game-changer in terms of scalability.

@Quentin_dQ @abhi1thakur The meilisearch vector method is experimental and didn’t work

@riadvargas @kiwicopple Amazing, thanks!

@Quentin_dQ @kiwicopple Great to hear that! My implementation is still based on pg_replicate (pre-renaming) so I will adapt it to new interfaces and loop you back

we are working on a Rust-based ETL server that can stream your Postgres database to S3/Iceberg (and other databases like BigQuery & ClickHouse) 100% open source, and designed so that it can be embedded in any Rust server

@kennyhyder @abhi1thakur Why do you use Qdrant and Meilisearch for the hybrid when Meilisearch will be more relevant by using the built-in hybrid search? Happy to help if needed

@abhi1thakur I’ve been struggling to get search working properly on marksearch.ai - currently struggling with a hybrid approach using Qdrant and Meilisearch. From first glance, Vespa seems like the perfect solution? Or is it overkill?

@riadvargas @kiwicopple I would be thrilled to see what you've done, and I would love to contribute by adding Meilisearch to the official destinations!

@kiwicopple Great tool! Have been using it for PG -> Meilisearch sync for a while and it has been surprisingly easy to operate

We just wrapped a huge week at Meilisearch with five major feature releases – chat with your data, geo-search on custom shapes, image + text multimodal queries, personalized results, and flexible usage-based pricing. We built these because search needs to evolve: to be faster, smarter, more intuitive, and fair from a cost standpoint. Thank you to our team and community – it’s only the beginning! I invite you to read the full recap of what we did and why → meilisearch.com/blog/launch-we…

Today we’re fixing the worst part of search pricing: anxiety 😬 We’re launching resource‑based pricing at Meilisearch! Our philosophy towards building Meilisearch has always centered on giving our customers maximum flexibility. That’s why we’re introducing a developer-first pricing model that stabilizes spend, aligns pricing with actual infrastructure, and provides clear controls over your performance and infrastructure. Just pick your compute (CPU/RAM). Storage auto‑scales. No caps on docs or searches. Predictable bills; direct control over performance. New Infrastructure tab to watch CPU, RAM, and storage in real time; alerts and recommendations up next. Want the full rationale, tiers, and how to switch? Dive into the full post on our blog meilisearch.com/blog/resource-…

🖼️ Day 3 → Enhanced Multi-modal Search Meilisearch now supports: ▪️ Image-based queries ▪️ Text + embedding combinations ▪️ Vector-native relevance No setup required – it’s built in. Demo → multi-modal.meilisearch.com Docs → meilisearch.com/docs/learn/ai_…

Day 2 of Meilisearch Launch Week 💥 Yesterday: Meilisearch Chat dropped – a chat interface to talk with your data. Today: it’s live on Product Hunt 👀 We built it because devs deserve search that feels conversational *and* runs fast ⚡️ Find us on the PH homepage and come say hi 👋

Launch Week starts NOW 🚀 Day 1: Meilisearch Chat – a chat interface powered by Meilisearch. Chat with your data. Search with context. ⚡️ More drops coming every day 👀 Subscribe to follow along → meilisearch.com/launch-week-q3…

@euro_rust And we wrap up with a great look into the journey of @meilisearch by Clément Renault (@kerollmops) , covering how they tackled the initial challenges, integrating semantic search, and then getting better at it with the magic of OSS (cc Nate Nethercott)!! 💫⚡🦀