AI Product Lead. Building MAX @Modular @Qualcomm. Mojo๐ฅ maximalist. Teacher at heart. Into powerlifting. Used to know some Math. Opinions are mine.
Vancouver, British Columbia
- Tweets 1,651
- Following 282
- Followers 623
- Likes 5,633
We're partnering with @Qualcomm to bring personal AI context to devices powered by Snapdragon. Liquid Context, our on-device context layer, is now optimized for Snapdragon processors and runs on the Qualcomm Hexagon NPU.
Our goal: Give the agents people choose an understanding of what matters to them and when they need help. With the user's permission, Liquid Context learns from device signals and builds an understanding of their routines, preferences, and needs. That understanding is built and maintained locally, and Liquid Context shares relevant context with the user's chosen agents, whether they run on the device, in the cloud, or across both. That includes third-party agents and Liquid Agent, our efficient embedded agent powered by LFM2.5-2.6B.
Running on the Hexagon NPU, Liquid Context works in the background and keeps that understanding current without requiring a cloud model to process every update.
For device manufacturers, this is a path to add personal context to their devices while supporting their own choice of agents and services. OEMs building embedded or hybrid agents can also work with us to evaluate Liquid Agent.
As our CEO @ramin_m_h said: "Personal AI starts with understanding how you live and what you need, when you need it. Liquid Context builds that understanding on your device so the agents you choose can offer more relevant help and anticipate your needs."
> Read more about our partnership: liquid.ai/blog/liquid-contexโฆ
> Check out the livestream of @cristianoamon and Ramin's keynote here: piped.video/watch?v=xKCto1Yfโฆ
19
60
400
37,096
๐๐ฅ๐ฐ๐๐ซ retweeted
.@clattner_llvm takes the #SnapdragonSummit stage for the first time as EVP Advanced AI Software @Qualcomm following the @Modular acquisition.
Introducing the audience to the stack and it's role in the broader ecosystem.
3
17
1,010
This is called the well-known fragmentation pain! Go and learn how folks deal with it while at @Modular we're solving it.
PyTorch releases are no longer just about shipping PyTorch binaries. Release engineering now coordinates PyTorch, Triton, and @vllm_project as part of a broader ecosystem.
At #PyTorchCon North America 2026, Andrey Talman (@Meta) will discuss how the PyTorch release process is being modernized, including the use of AI agentic workflows to analyze failures and the work required to release projects together with reliability and performance in mind.
Hear more from Andrey about what heโll cover at the conference.
๐ Register: events.linuxfoundation.org/pโฆ
1
8
390
One of the outcomes from AI in math and autoformalization that I'm looking forward to is showing the gaps and incorrect theorems. In my grad school, I learned that requiring 100% bullet proof results isn't practical which wasn't the type of math I wanted to exist, but now I can!
2
139
This only shows the math community needs to grow up and take the hard reality. Everything with math has changed and they need to adapt. Chess-players witnessed an early version of this, then Go, then CS and programming and now math. It's adapt or become extinct situation.
Twenty-five Fields Medal winners have published a joint declaration warning about what they see as a severe misalignment between AI companies and the mathematics community.
8
513
Today's AI software is fragmented. Every accelerator has its own compiler, kernel libraries, runtime, and development path. Developers absorb the cost of this fragmentation.
In his ModCon 2026 tech talk, Abdul Dakkak, Chief Scientist at Modular, presents our alternative: a unified compute layer, flexible enough to extend to new models, modalities, and hardware.
Abdul walks through our recipe for bringing up new hardware, demonstrated with three examples:
1. AWS Trainium, via our own team. Gemma 4 31B end to end.
2. Google TPU v6e, through our partner @HTECgroup. Their team brought it up with no LLVM backend to lower to, no prior knowledge of Mojo or MAX internals, and minimal support from us.
3. d-Matrix Corsair, implemented by @dMatrix_AI's own team on their own stack. Repo access Tuesday, working matmul by the following Monday.
Thanks to HTEC and d-Matrix for their collaboration, and to Mihailo for presenting HTECโs learnings.
Read about the HTEC collaboration: htec.com/insights/media-coveโฆ
Watch the full talk: piped.video/watch?v=hYjKbTAGโฆ
1
5
22
1,643
๐๐ฅ๐ฐ๐๐ซ retweeted
not lost on me that the two labs asking the public to trust them as stewards of responsible AI leadership at existential stakes are having a slapfight on twitter about who gets credit over a math problem
17
100
934
62,512
๐๐ฅ๐ฐ๐๐ซ retweeted
I resigned from Anthropic today. I spent the last three years doing pretraining research at both OpenAI and Anthropic. Neither company is acting responsibly. They are racing straight to self-improving superintelligence and gambling with our lives. More thoughts below.
20,633
167,475
803,100
174,276,771
The gravity of this is so much, it'll take quite some time to sink in!
Weโre sharing a solution to the Navier-Stokes Millennium Prize Problem, one of the deepest problems at the frontier of mathematics.
The proof was produced by a group of agents, using an OpenAI next-generation model significantly more capable than GPT-6 Astra.
The problem concerns whether the description of smooth three-dimensional fluid motion modeled by the Navier-Stokes equations can break down. It has remained unresolved for roughly 90 years.
1
209
MAX is Modular's inference stack for running any model on any chip: one codebase, one model definition, one kernel library. High-performance AI serving and modeling for any hardware.
In this technical deep dive from ModCon 2026, Senior AI Product Manager @ehsanmok and Head of MAX/Senior Director Bingfeng Xia open up the stack layer by layer and show the code behind each claim:
piped.video/3H8Orjaa-bk
4
29
2,646
๐๐ฅ๐ฐ๐๐ซ retweeted
Caught up with @clattner_llvm right after ModCon. Expanded hardware support, Mojo now open source, Modular Cloud live, and a new industry alliance. As he put it: heterogeneous compute now has a software platform.
2
11
67
7,305
Our #ModCon26 talk with Bingfeng Xia is now available!
Learn about the internals of MAX, the next generation AI framework we're building at @Modular
piped.video/watch?v=3H8Orjaaโฆ
1
500
Modular's price-performance on @Zai_org's GLM-5.2 (Non-reasoning) lands right on the Pareto frontier in @ArtificialAnlys' latest benchmark: near-top speed without the near-top price tag.
We're just getting started, and we're ready for GLM-5.3. Expect to see a lot more incredible results. ๐
2
11
102
7,951
๐๐ฅ๐ฐ๐๐ซ retweeted
My take on @Modular's ModCon 2026 I attended Tuesday. "Modular Delivers Openness And Accelerator Portability At ModCon 2026"
forbes.com/sites/patrickmoorโฆ
8
32
21,018
๐๐ฅ๐ฐ๐๐ซ retweeted
ModCon '26 is a wrap! Thank you to everyone who attended + we will post videos for folks who couldn't.
The theme: "Heterogenous compute now has a software platform, and it's real".
Modular and QCOM delivered on our promise: we open sourced Mojo๐ฅ, added new hardware, and more!
15
12
183
13,434
Today, we open sourced Mojo ๐ฅ.
Announced just now during the ModCon keynote, effective immediately, Apache 2.0 License.
Thank you to our community for waiting patiently and building alongside us.
#ModCon2026
Full blog: modular.com/blog/mojo-open-sโฆ
47
225
1,414
96,441
A highlight of my career is to be able to see this!
#Mojo is completely open sourced, Apache 2.0 ๐๐๐
1
10
731
๐๐ฅ๐ฐ๐๐ซ retweeted
Modular CEO @clattner_llvm kicks off ModCon 2026. He starts off with a core value prop review followed by some huge new support announcements:
- TPU
- Trainium
- Qualcomm
Adds to NVIDIA B200 and MI355x.
And MOJO is open-source. This could be the biggest announcement.
$QCOM
6
9
55
5,745
ModCon 2026 nitter.net/i/broadcasts/1NGarozOnโฆ
3
11
51
338,352