English

Posts with #ModularAI

#ModularAI

Public posts on X that use #ModularAI.

Related hashtags

  1. kirill@kir307
    @kir307
    What about several small models instead of one large model, each with its own job? Each has its own weights. The models I need won't all fit in 16 GB. Swapping them in and out means waiting. Passing results through generated text and JSON adds latency. #LLM #AIEngineering
    1014
  2. kirill@kir307
    @kir307
    To make AI faster, remove what the task doesn't need: extra weights in memory, intermediate text generation, network calls, repeated processing. Keep language understanding in one shared core. Add only the skills needed. That's how I'm building ModularAI. #ModularAI
    10110
  3. kirill@kir307
    Can we solve LLM tasks in tens of milliseconds on a Mac mini M4 with 16 GB? I set out to prune experts from an MoE model. I ended up with a different architecture: a shared core and neural skill packages. Here's why—and what I'm building. #ModularAI #LocalAI
    10020
  4. kirill@kir307
    @kir307
    ModularAI's goal: solve practical tasks in tens or hundreds of milliseconds on a Mac mini M4 with 16 GB. So AI can run right inside a backend request: call it, get a result, keep going. A regular part of a microservices architecture. #Microservices #dotnet
    1009
  5. kirill@kir307
    @kir307
    ModularAI has big ambitions—and a lot of work already done. This is well beyond an idea on paper. I'll share the research, what works, and what I've had to rework. Working demos are coming soon. #ModularAI #BuildInPublic
    0009
  10. 𝔈𝔥𝔰𝔞𝔫✓@ehsanmok
    Let's gooo @Modular! Truly native GenAI stack built on Mojo🔥 with Python 🐍 API that runs on GPU i.e., no fragmentation and NO C/C++/CUDA or anything else involved, and that's just the start 🤯 #genai #ModularAI #gpu #native
    Modular @Modular
    MAX 24.6 is here and available for download today! Say hello to MAX GPU: the first vertically integrated gen AI serving stack, with SoTA performance on NVIDIA A100 and support for deployment across all major clouds 🤯 Don't miss the announcement:
    106482
  11. Modular✓@Modular
    MAX 24.6 is here and available for download today! Say hello to MAX GPU: the first vertically integrated gen AI serving stack, with SoTA performance on NVIDIA A100 and support for deployment across all major clouds 🤯 Don't miss the announcement:
    116105286K
  14. GodAir✓@dang_duytan
    Everyone’s building L2s. Few are building L2s for agents. @tenprotocol is laser focused: AI native infra, optimized for autonomous agents to transact, coordinate, and scale without relying on outdated Web2 bridges. This is serious infra. #AgentEconomy #modularAI #tenprotocol
    Download
    20215235.1K
  22. Prieth@PriethOnchain
    🎮 Most games: – You play. – You grind. – You own… nothing. Enter @DubheEngine on Sui 🚀 AI-driven. Modular. Player-owned. Build smarter worlds where assets live on-chain & players shape the game. The future of gaming isn’t Web2. It’s Dubhe. #SUI #ModularAI
    Download
    106144
  24. Show more posts