Pension_qe

GitHub:

[Pension_qe @Pension_qe]

Meet Isvik.cpp.

A C++20 local AI runtime for on-device inference.

Multi-backend support:
• OpenVINO GenAI
• ONNX Runtime GenAI
• TensorRT
• TensorRT GGUF (experimental)
Run local AI with desktop chat, CLI, and API services across CPU, GPU, and NPU.
https://t.co/zIFUQysOcY
#LocalAI #OpenVINO #ONNX #TensorRT

[Pension_qe @Pension_qe]

Just built a Nim API for OpenVINO Runtime 🚀

It wraps the OpenVINO C API with both low-level bindings and a more idiomatic managed Nim interface.

Windows + Linux tested.

GitHub:
https://t.co/WCYbeCmGfY

#NimLang #OpenVINO #AI #OpenSource

[Pension_qe @Pension_qe]

Intel's strange peripheral products socks

[Pension_qe @Pension_qe]

与我一起赚取 500 Microsoft Rewards 积分! https://t.co/kpRGyZGc2b #MicrosoftRewards 来自 @MicrosoftEdge

[Pension_qe @Pension_qe]

@X_supporttop1 Yse

[Pension_qe @Pension_qe]

@learning_yohei 🇹🇼

[Pension_qe @Pension_qe]

Shanghai

[Pension_qe @Pension_qe]

LOL

[Pension_qe @Pension_qe]

@samueljmcd AI

[Pension_qe @Pension_qe]

I just reserved my Cloudflare Wallet tag: https://t.co/3FkdVYiDwv.

Reserve yours now at https://t.co/YYUGDwVEg9

[Pension_qe @Pension_qe]

🚀 Introducing Isvik — a native local AI inference runtime for Intel AI PCs.

Built with 100% pure Nim and directly integrated with the OpenVINO C API.

• CPU / Intel iGPU / Intel NPU
• Zero Python runtime dependency
• Single-file native binary
• OpenAI & Anthropic compatible APIs
• Local-first, no cloud inference
• Apache 2.0

Isvik is designed to make local AI inference feel like a native part of the PC — not another Python stack.

🌐 https://t.co/5IV8jOwqBN
💻 https://t.co/7CQhUwf1Dk

[Pension_qe @Pension_qe]

Isvik is an open-source, high-performance native local AI inference runtime built specifically for Intel AI PCs.
Written 100% in pure Nim with zero Python runtime dependency, it talks directly to the OpenVINO C API and delivers deeply optimized inference across CPU, Intel iGPU, and NPU.
Key advantages:

Single lightweight binary with sub-100ms cold starts
Intelligent automatic device scheduling (NPU → iGPU → CPU)
Native support for OpenAI Chat Completions, Responses API, and Anthropic Messages
Clean interactive workbench + full local API server
Industrial-grade safety and honest benchmarking

No bloated environments. No unnecessary dependencies. Just fast, native local AI.
Fully open source: https://t.co/7CQhUwf1Dk

[Pension_qe @Pension_qe]

@samueljmcd I like Claude Opus 5 and Gemini 3.1 Pro

[Pension_qe @Pension_qe]

The ultimate AI workflow for devs: "Sleeping Agent" 😴🤖
GPT-5.6 Luna is pretty slow. But that’s actually a feature, not a bug.
When running huge refactors or test suites, sitting around waiting is painful.
My Nightly Setup:
1. Brain: Plan architecture & test cases before bed.
2. GPT-5.6 Luna: Grinds through the auto-loop (Write → Test → Fix) at its own pace overnight.
3. Morning: Wake up and directly inspect the completed project.
It takes its time, but you’re asleep anyway. 🔥
What tasks are you delegation-dumping to Luna overnight?

[Pension_qe @Pension_qe]

Don't give the AI a whole set of prompts.