
Jaana Dogan ヤナ ドガン
@rakyll • 171,048 subscribers
ML/AI Systems at Google. Simplicity and optimism. Personal opinions.
Shorts
Videos

🥳 Excited to start revealing what we've been working on in the last few months. First, we decided to reinvent Kubernetes for agentic workloads with statefulness and fast resumption. Secondly, we are building an agentic orchestrator that will be Google's open agentic orchestrator and runtime.
Jaana Dogan ヤナ ドガン724,244 Aufrufe • vor 1 Monat

AX Studio started as an afternoon prototype but I'm thinking we should build this as an official tool. - Start any task from prebuilt images: Gemini, OpenAI, Clade agents - Manage and inspect tasks, state, logs - SSH into tasks (if they allow) - Automatically setup clusters
Jaana Dogan ヤナ ドガン14,619 Aufrufe • vor 11 Tagen

Two days ago, I was thinking that we made tooling hard for AI developers who need containers. I built aic, which works like Docker (for now) but will explore composition of images soon. Last night, generated code started to work. It runs things in isolation and copies the workspace etc.
Jaana Dogan ヤナ ドガン20,543 Aufrufe • vor 18 Tagen

We spent the last few months building an incredibly fast resumable stateful sandbox layer. We can suspend and resume so fast that you can do it between tool calls without impacting user experience. Each command in the video is resuming and suspending the underlying sandbox.
Jaana Dogan ヤナ ドガン57,196 Aufrufe • vor 3 Monaten

Last night, I rebuilt my earlier Google office simulator with Gemini Flash 3.7. It took me 7-8 shots to get to what you see in the cast, in comparison to 50+ shots with a pro model 18 months ago. The leap between March 2025 and August 2026 is impressive.
Jaana Dogan ヤナ ドガン38,926 Aufrufe • vor 2 Monaten

Faster and higher quality models are probably going to emerge their own use cases. We were playing around with the idea of using Gemini Flash 3.7 for orchestration. It's NOT intended for production use, but writing Kubernetes specs in natural language is viable.
Jaana Dogan ヤナ ドガン13,314 Aufrufe • vor 2 Monaten

Multimodality and streaming is hard. I've been building something that allows you connect streaming devices like screen capture, microphone, camera, and text easily to craft generative streaming pipelines. It works well with Gemini 2.0 models. Happy to open source if anyone wants it.
Jaana Dogan ヤナ ドガン17,851 Aufrufe • vor 1 Jahr
Keine weiteren Inhalte verfügbar