Jason Cui

🚀 Excited to share that I’ve joined @a16z as a partner on the infrastructure team, where I’ll be investing in all things infra, data, and AI.

🚨 New: Excited to announce we @a16z are leading @deeptuneai’s Series A. RL environments are becoming both the bottleneck and the unlock for training the next generation of frontier AI models. The shift is clear: from static datasets to dynamic, engineered environments where models learn to write code, perform knowledge-work tasks, and fully control computers end-to-end. Deeptune is building this layer from the ground up and is already driving progress with top AI labs. Learn more about Deeptune and @timlup:

@Mascobot @a16z @deeptuneai 🔥🔥🔥

We've reached an agreement to acquire Astral. After we close, OpenAI plans for @astral_sh to join our Codex team, with a continued focus on building great tools and advancing the shared mission of making developers more productive. openai.com/index/openai-t…

@marty_kausas No more purple building??

We're moving out of the Pylon building! We moved to this office in July 2024. We were 13 people at the time. Over the next 18 months we'd grow to 90 people, ship endless product, build new functions from scratch, and start serving thousands of customers.

@KenMFLau 💯

@JasonSCui fine fine fine let’s put up a bid together

What a modern living community in America can and should look like

@marty_kausas You get incredible ear-to-the-ground exposure by building everyday - by chatting with teammates, other founders, customers, etc

Love this "overnight success" story. The founders of Moltbook spent 10 years building another startup, decided to hard pivot, went viral, and then got acquired 3 months later by Meta. -- This is not just common, it's the norm. Most founders I know spent years working on things that didn't go anywhere, and then suddenly things start working. This was once described to me as "you move slow, and then suddenly move fast." Same thing happened to us at Pylon. I spent 2 years pivoting across ideas and industries with multiple different co-founders before we founded Pylon. Robert and Advith separately spent 1.5 years working on things that didn't work. Then suddenly, 2 years in, things started working. -- The Moltbook founders took this to the extreme working on another company for 10 YEARS that was probably doing ok, but took a risk on an incredibly risky pivot. I've never met them but congrats to the Moltbook team! Very cool story and massive respect.

@MichelleLazzar 😂 will need more than a drop!

AI models to the data layer

@DBredvick @turbopuffer @NotionHQ @youngbloodcyb Love it!

@JasonSCui We built this all in-house at Vercel on top of @turbopuffer + @NotionHQ. I wish there was something off the shelf — hope your call-for-startups works! @youngbloodcyb built our context layer and we're getting it plugged in to more agents. ROI has been great.

Frontier models are exceptionally efficient, intelligent, and useful. For agents, context is now the bottleneck. Enter the context layer, which bridges the gap from an enterprise's messy data to actionable context, packaged for agents. We're seeing three distinct verticals emerge in the context layer space: - Data gravity platforms - Existing AI data analysts - New, dedicated context layer companies Read the full piece by @JasonSCui and @JenniferHli: a16z.news/p/your-data-ag…

@inferact @nvidia @vllm_project 🚀🚀🚀

We are thrilled to announce that @nvidia is the latest investor in @inferact. We look forward to continuing the momentum driven by our deep collaboration: (1) Engineering velocity: a significant uptick in @nvidia pull requests to the @vllm_project repo. (2) Product synergy: close integration with NVIDIA Dynamo, ModelOpt, Nemotron, and more products! It’s an exciting time for the growth and development of vLLM, the world's AI inference engine!

@JenniferHli Context is the bridge! 🤝

The main obstacle between today's frontier models and working agents is context. It is certainly the case for data agents too. For the first time we're getting really close to an AI analyst by your side that can answer any question, at any time. @JasonSCui and I wrote about context layer, the critical piece that which is bridging the gap between the messy enterprise and the packaged data agents need:

@BlasianHokage All it takes is one effort to deploy an enterprise agent not grounded in the right context to feel the pain

@JasonSCui Good piece. Suprised why everyone is always so focused on enterprise context. Good news for me I guess

Super interesting work from team Databricks! Some interesting tidbits: - KARL is cheaper per query than its own base model (GLM 4.5 Air). RL efficiency gains! - They trained compression end-to-end via RL so the model learned what to forget to max reward - PMBench is the internal proprietary eval over internal PM meeting notes. This was the hardest task - can confirm as in my prev product role we took product notes seriously and there a massive repository of them :)

Crazy what you can accomplish with long running agents from your phone. Great process documentation by @BornsteinMatt!

@SahilBloom @levelsio Why no whey?

@levelsio It’s a full one of these small sized cartons. 50g protein. Honestly, if I didn’t have time, I’d just drink it.

Unbelievable the amount of food you need to eat to hit 2g per kg in bodyweight while being in caloric restriction or maintainance, I mean I hit it but it's not easy is it

@BornsteinMatt @mlg27_ Congrats Michael this is awesome!

this is a genuinely useful app: text to design. amazing what @mlg27_ has done as a single dev still in high school (middle school??)😂

It's a pleasure to work with the @SAIRfoundation as they bring together the community around AI for Science. Lots of exciting things to come!

Muji pens are special, love them as my daily driver. Mine is 0.38mm though, also with a cap :)