π”ˆπ”₯π”°π”žπ”« (@ehsanmok) header
π”ˆπ”₯π”°π”žπ”« (@ehsanmok) profile picture

AI Product Lead. Building MAX @Modular @Qualcomm. MojoπŸ”₯ maximalist. Teacher at heart. Into powerlifting. Used to know some Math. Opinions are mine.

Vancouver, British Columbia
Joined July 2014
Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)Photo by π”ˆπ”₯π”°π”žπ”« (@ehsanmok)
π”ˆπ”₯π”°π”žπ”« retweeted
44
353
29
4,412
78,949
π”ˆπ”₯π”°π”žπ”« retweeted
MNIST of 2027
holy fucking shit Levent AlpΓΆge released the Lean code for his claimed proof of the Riemann Hypothesis without releasing a human-readable paper because the write-up is over 250,000 pages long
5
9
1
166
7,522
π”ˆπ”₯π”°π”žπ”« retweeted
We redesigned MAX’s Module API to make building fast models even faster. It’s PyTorch-like, so the syntax feels familiar. It also runs eagerly, so you iterate and debug without waiting on compilation. When you’re ready for production, one compile call gets you the full MAX graph compiler. Try the new Module on your next model, and explore the full improvements: modular.com/blog/reintroduci…
10
51
3,603
I'm very excited about Mojo's comptime reflection making Rust's proc macros unnecessary! Watch @clattner_llvm SPLASH'26 talk on Mojo: Towards One Systems Language for All Hardware youtube.com/watch?v=9-97SB5X…
4
302
π”ˆπ”₯π”°π”žπ”« retweeted
MNIST of 2026
Replying to @claudeai
A team of 1,000 Fable 5.2 agents fully formalized Fermat’s Last Theorem in 30 hours, translating the proof into a language a computer can check step by step. With the same setup, Opus 5.5 completed only 26% in 18 hours, and barely made any more progress after 30 hours.
6
14
6
286
10,772
π”ˆπ”₯π”°π”žπ”« retweeted
Super excited about our new license, come build with us! πŸ”₯πŸ™πŸΌπŸš€
2
1
6
346
π”ˆπ”₯π”°π”žπ”« retweeted
Wondering what the new MAX License means for your project? @clattner_llvm and @iamtimdavis are taking your questions tomorrow at 10 AM PT. Build, modify, and ship with MAX, and get the details from Chris & Tim: luma.com/max-license Submit your questions here: forms.gle/kYzFgDCgsaqZQPqD8
2
6
18
1,909
π”ˆπ”₯π”°π”žπ”« retweeted
LFM3 from @liquidai is now running on MAX. πŸ§‘β€πŸš€ All 6 models in the suite are supported, across text, audio, vision, and omni. The implementations are open source, so you can read the code, run it, and build on it: max.modular.com/api/python/p… Which modality are you trying first: text, audio, vision, or omni?
Introducing π™»π™΅π™ΌπŸΉ: our new generation of Liquid Foundation Models. With a new tri-hybrid architecture and new multimodal encoders, the LFM3 family is the ideal model suite for specialized intelligence. > 6 model instances, across text, audio, vision, and omni modalities, generative and decision models. > π™»π™΅π™ΌπŸΉ-𝟷.πŸ½π™±: text, 256k context window. > π™»π™΅π™ΌπŸΉ-πš…π™»-πŸΈπ™±: vision-language model with video-input support for spatial understanding > π™»π™΅π™ΌπŸΉ-π™°πšžπšπš’πš˜-πŸΈπ™±: tool calls, voice chat, reasoning in 18 languages. > π™»π™΅π™ΌπŸΉ-π™Ύπš–πš—πš’-πŸΈπ™±: an omni model combines vision, audio, and language. > 𝚍𝟷-πŸΈπ™± and 𝚍𝟷-π™Ύπš–πš—πš’-πŸΈπ™±: decision models with typed outputs and calibrated probabilities in a single pass. Omni adds audio input. > fine-tune on 1 GPU. > a new tri-hybrid architecture > architecture optimized for 14 processor classes across GPUs, CPUs, and NPUs. > π™»π™΅π™ΌπŸΉ-πš…π™»-πŸΈπ™± is a beast. It performs competitively with Gemma 4 12B (6x larger), and Gemini 3.5 Flash Lite, and GPT-5.6 Luna, with 1.9B parameters. > π™»π™΅π™ΌπŸΉ-𝟷.πŸ½π™± is #1 on the Pareto frontier of device intelligence πš™πš’πš™πšŽπšπšπšŽ.πš•πš’πššπšžπš’πš.πšŠπš’ Every modality for specialized intelligence. Pick the one that fits your use case, customize it, and run it on-device. 1/ 🧡
2
5
1
47
9,360
π”ˆπ”₯π”°π”žπ”« retweeted
Liquid 🀝 @Modular ✨
LFM3 from @liquidai is now running on MAX. πŸ§‘β€πŸš€ All 6 models in the suite are supported, across text, audio, vision, and omni. The implementations are open source, so you can read the code, run it, and build on it: max.modular.com/api/python/p… Which modality are you trying first: text, audio, vision, or omni?
7
117
7,569
π”ˆπ”₯π”°π”žπ”« retweeted
finally found a company that truly listens to its customers
77
254
44
4,363
145,098
Flare v0.13.0 is the first and fastest #MojoπŸ”₯ networking library that is ranked high all across in HttpArena http-arena.com/#scope=h2&typ… Checkout github.com/ehsanmok/flare/re… A blog is coming!
1
1
13
743
π”ˆπ”₯π”°π”žπ”« retweeted
Bonsai 2 2B running at 40 tok/s on Snapdragon 8 Elite Gen 5, entirely using the Modular stack. No llama.cpp! It was surprisingly easy to add Android as a target to Modular. And I think this bad boy can go faster!
I'm happy to share that Mojo and MAX are now fully open - for development, extension, and commercial use! This includes the full Mojo programming language, but also the MAX graph compiler, MLRT runtime internals, and hardware abstraction logic. Get your agents going!πŸ‘‡
6
11
1
106
5,827
π”ˆπ”₯π”°π”žπ”« retweeted
I'm happy to share that Mojo and MAX are now fully open - for development, extension, and commercial use! This includes the full Mojo programming language, but also the MAX graph compiler, MLRT runtime internals, and hardware abstraction logic. Get your agents going!πŸ‘‡
49
192
23
1,505
79,690
π”ˆπ”₯π”°π”žπ”« retweeted
MiniMax M3.1 Flash (aka Space Bunny) from @MiniMax_AI is live on Modular Cloud, day zero. M3.1 Flash is best in cowork automation among open models, with SOTA-level coding and engineering performance. It can ship a fix in a real codebase, then update Jira and Slack afterwards. We worked with the MiniMax team ahead of launch, so M3.1 Flash is already serving production traffic on Modular Cloud. Evaluating it for your stack? Try it now: console.modular.com/signup?u…
17
5
1
51
5,716
π”ˆπ”₯π”°π”žπ”« retweeted
Also in 26.7: eager mode for faster custom op development, decision model support, and Mojo 1.2 with pattern matching, the next step in Mojo’s growth into a systems programming language: modular.com/blog/26-7?utm_so…
1
1
4
601
π”ˆπ”₯π”°π”žπ”« retweeted
The Modular alliance program extends that foundation, giving hardware vendors and partners a shared path to build on MAX together: modular.com/alliance?utm_sou…
1
3
12
2,272
π”ˆπ”₯π”°π”žπ”« retweeted
The new MAX License balances two priorities: freedom to build, and a unified, sustainable ecosystem. Anyone can use, modify and ship commercial products with MAX, without licensing fees or royalties. Its few restrictions prevent fragmentation into incompatible versions, so everyone builds on a common foundation. Full details on the license: modular.com/blog/max-license…
1
1
8
709
Source availability of MAX is an important step towards more open and developer-first platform.
AI shouldn’t be locked to one hardware vendor or chip. To democratize AI compute, we’ve made MAX source available with Modular 26.7. 🧡
1
226
No understanding means no taste. No taste means no direction. No direction means one is lost!
1
1
5
284