Sabitlenmiş Tweet
Steeve Morin
26.3K posts
Steeve Morin
@steeve
Building @zml_ai (and we're hiring), ex @zenly, ex Exalead, ex @google. Skydiver and wingsuiter.
Paris, France Katılım Ekim 2010
1.3K Takip Edilen9K Takipçiler
I’ll be in SF Bay Area next week and will be attending @AMD @AIatAMD conference with @corentinanjuna.
Hit me up if you want to chat on new silicon, very high performance inference and Zig.
Also I’ll be demoing ZML/LLMD at the meetup organized by @andrey_cheptsov on the 23rd.
English
BuildBuddy is literally magic. If I ran the stuff it gets done in minutes on my laptop, it would explode.
English
Je ne vais pas vous cacher à quel point je suis fier de celle-ci. Et un peu ému aussi.
Voici le nouveau véhicule du Président de la République française, présenté aujourd’hui lors du défilé du 14 juillet, et ses suspensions (oléopneumatiques) sont des @DriveNimbus.
Je ne pensais pas vraiment, quand nous avons créé Nimbus il y a un peu plus de 3 ans, que nous aurions l’honneur de contribuer à la mise au point de ce véhicule d’exception pour un passager aussi prestigieux.
Je pèse mes mots et je ne m’enflamme pas. Nous ne sommes toujours personne sur le marché mondial, et ce ne sont que des suspensions. Mais il se trouve que les suspensions, c’est ce que nous faisons, donc être retenus pour ce projet dans ce qui est notre spécialité est une victoire en soit pour notre petite équipe.
Bravo à tous les Nimbusiens qui ont participé à cette aventure, je suis vraiment vraiment vraiment fier de vous.
Français
Here is your chance to meet someone exceptional
Corentin Kérisit@corentinanjuna
I'm spending 1 month in Tokyo. Anyone I should meet ? -- Startups, Developers, founders, expats, friends of friends ? 東京に1か月います！ この機会にぜひいろんな人と会いたいです。 スタートアップ、エンジニア、起業家、海外勢、友達の紹介などあればぜひ！
English
Steeve Morin retweetledi
‼️ BREAKING: xAI's Grok Build CLI was uploading entire Git repositories to a Google Cloud bucket, private codebases and unredacted secrets included. The uploads quietly stopped via a hidden server-side flag, and xAI still has not said a word about scope, retention, or deletion.
The scale is staggering. On a 12 GB test repo, 5.1 GB flew out the door to xAI's grok-code-session-traces bucket while the actual coding task needed just 192 KB. The tool grabbed whatever repository it ran in, not the files it needed.
The fix arrived as a hidden flag, disable_codebase_upload: true, a day after a researcher's wire-level analysis. The "Improve the model" opt-out never stopped the uploads.
Still no advisory, no scope, no word on whether already-uploaded code gets deleted. For anyone pointing AI coding agents at proprietary code, what crosses the wire matters more than what the settings page says.
English
@P_Bonnet @ClickHouseDB i'm a simple man, I see clickhouse, I like
English
I think you forget @ClickHouseDB in your conclusion. Best placed for an agentic world imo (instantaneity and cost are two of its features which will be most desirable as agents take work off humans: many more pulls from DB, and can consume instantly).
Asymmetric Bets@UncleAlpha007
Satya Nadella’s "Reverse Information Paradox" meets Gavin Baker’s CapEx thesis The Big 3 hyperscalers are trying to build chips to replace Nvidia. So Nvidia is using open-weight models like Nemotron and Sovereign Neoclouds like Boost Run $brun to bypass the hyperscalers entirely. If you want to own the enterprise AI stack long-term, you don't rent general-purpose brains—you own the sovereign data vault. The companies with the biggest revenue inflection from this are, in my view, $Brun and $nbis. Longer term huge winners are likely to be $snow and databricks.
English
@penberg do you know if there are drives that are announced on that front ?
English
@steeve Not sure i am answering your question, but I believe Optane essentially killed NVMM and CXL-SSD is the successor
English
what is the spiritual successor to Optane? meaning very very low latency flash
XL-FLASH is ""okay"", Z-SSD is discontinued...
English
I made the grave mistake of accepting a teams call, and now I can't log into the org of another company we share a channel with. 3hrs trying so far.
English
@hardfist_1 You are right. We are a complez scenarios involving FFI codebase and memory management is not hard in Zig imho
English
I like Zig too, but I’m not convinced by the argument that TigerBeetle proves memory management is not hard in Zig, and that projects struggling with it simply have a “skill issue.”
TigerBeetle’s memory model is fundamentally different from that of a typical application. Many real-world workloads are difficult, or even impossible, to estimate in advance. For something like the Bun bundler(or Rspack), which has to process arbitrary user code, I don’t see how you could reliably predict an upper bound on memory usage ahead of time.
So we may need a better example(ghostty maybe a better one) of how memory can be managed well in a large, general-purpose system—especially in complex scenarios involving FFI, like Bun. TigerBeetle is an impressive project, but it is not a representative case.
English
English
English
More and more are realizing
Corentin Kérisit@corentinanjuna
Have you ever stared into the abyss as Rust "aws-lc-sys" build script took an eternity? Well, we have great news for you - We now have fully bazelized "aws-lc", fully hermetic bindgen and everything "just works". The best part of all? Just use/update rules_rs, run a build, and it will tell you exactly what to do. Check out rules_rs if you want fast hermetic builds on easy mode, curated by experts 🔥 @bazelbuild keeps proving how much of an unfair advantage it is...
English
@1984perdra @zml_ai non, tu imagines bien que je decouvre pas rust
Français
@harderthanfire @zml_ai the general design allows for incredible optimizations and control
explicit allocators and io come to mind but it's a general thing imho
which is maybe it's also hard to explain to others that it's not about "yeah it's LLVM so it's all the same"
English
What would you say the biggest benefit you've gotten from choosing Zig over any other low level language?
Personally I use Zig because it makes it much easier to reason about lifetimes, memory, cache usage etc making it great for data driven design vs some other lower level languages (yes C can do it too but not as nicely).
English
