Murat Sari

@wolfmanfx.bsky.social

Big Angular + AI Meetup in Graz Join us on September 11, 2026 for an afternoon packed with talks on Angular, AI, Agentic UI, and modern frontend architecture. 🎤 10+ international speakers 🍕 Free food and drinks Participation is free, but seats are limited. Link in comments 👇

Bild

What are Claude Code & Codex doing under the hood? At a minimum: Running a model, a loop, and a few tools. @wolfmanfx.bsky.social explains how to build a fully local coding agent from scratch: llama.cpp, a hand-written harness, and NVIDIA OpenShell as the sandbox. soverius.ai/blog/impleme...

Build a local AI coding agent from scratch

Build a local AI coding agent from scratch — Gemma 4 on llama.cpp, three tools, one loop — then learn why running it unsandboxed is dangerous and how NVIDIA OpenShell contains it.

soverius.ai

For a long time, AI discussions were mostly about the model. 🤖 Which one is better? 📊 Which benchmark was beaten? 🚀 Which new release changed everything? But only the model is not enough. The layer around it is just as important That's the harness. 📖 soverius.ai/blog/what-is...

What is a Harness?

Learn what an AI harness is, why it matters, and how it improves LLMs through tools, context management, agent runtimes, guardrails, and intent alignment.

soverius.ai

AI offers massive potential for UIs. The interface can be dynamically generated for each user. A key technology is #A2UI, which is framework-agnostic. Our new post covers: ✅ What is A2UI? ✅ How does it work? ✅ How to run it fully autonomously. soverius.ai/blog/behind-...

A2UI, Generative UI, LLM Benchmark, Local LLMs, DGX Spark, vLLM, llama

Can local LLMs produce valid A2UI? I benchmarked 14 model and engine combos on a DGX Spark. The 14B cliff and the failure modes are not what you'd guess.

soverius.ai

Two weeks ago at AI India, I gave a talk covering the core fundamentals of LLM Models and their current limitations. Here's the link to my slides. I’m also working on a detailed article covering these topics, which will follow soon. soverius.ai/talks/demyst...

Demystifying the AI Jungle: Connecting the Dots

Open-weight models, fine-tuning, skills... the list of AI buzzwords grows every day. When I first started digging in, I had the impression that a new technology popped up every month just to replace t...

soverius.ai

GPT 5.5 has doubled its prices compared to GPT 5.4. GitHub Copilot just switched to token-consumption models, and Opus 4.7... well 🤷. The era of "subsidized AI" is ending. Vendors need to earn money, and I expect prices will increase even further in the short term. soverius.ai 🧵 1/6

Soverius AI — Build Real AI Skills. Ship Real Products. | Soverius AI

Hands-on TypeScript & PyTorch workshops for developers and teams. Build production RAG pipelines, ship AI agents, and master the Claude & OpenAI APIs.

soverius.ai

Ok, this is huge 🤯 My colleague Murat Sarai aka @wolfmanfx implemented cool new code analyses for @softarc/detective. Display your modules, domains, etc, as a 3D city (credits to @richardwettel's CodeCity) 1/7

Bild

We just hit a milestone in our exploration of signal-based forms in Angular - our prototype is now merged into the main repository. This opens the door for some important next steps like compiler integration & interop w/ existing forms.

Writing code in Angular templates just got EVEN BETTER 🔥 You can now use untagged template literals expressions. Available now in Angular 19.2

✨Heads up! Observables are shipping natively to a browser near you soon! Huge thank you and congrats to @domfarolino.com ❤️ I've helped / worked on / watched this through different standards bodies for more than 10 years now. ☠️ (I'm going to have a lot of work now to adapt RxJS to this 😅)

BildBildBildBild