Vitalij Neverkevic Blog

Vitalij Neverkevic Blog

Latest — Jun 20, 2026

Dude, where's my GPU?

The global hardware and semiconductor market in 2026 is currently experiencing a profound and persistent structural imbalance. Driven by the collision of unprecedented infrastructure demand for artificial intelligence (AI) and severe geopolitical supply chain constrictions, the environment has evolved from a cyclical semiconductor shortage into a systemic crisis affecting every

More issues

The Axios Supply Chain Attack

Why This Was a Wake-Up Call — And What It Tells Us About Trust in Modern Software The software supply chain has evolved into the primary theater of operation for advanced persistent threats, and last night, we witnessed a nuclear-grade bomb drop on the JavaScript ecosystem. Let us not

The State of AI Agents (2025): Between Hype, Hard Truths, and Real Progress

TL;DR Everyone’s talking about “AI agents”, autonomous digital workers that were supposed to plan, decide, and execute on their own. But as 2025 winds down, reality is setting in. Most companies that tried full autonomy hit the same walls: errors that compound, bills that explode, and systems that

Understanding Fine-Tuning Challenges in LLMs (QLoRA, LoRA, and Beyond)

Why read this? Because the way we fine-tune LLMs (QLoRA/LoRA) might be quietly breaking the very intelligence we paid for... and the fixes aren’t what you think. TL;DR * Your model forgets... on purpose. Fine-tuning carves a deep, narrow canyon in the weight landscape. Great at

Microsoft Shakes Up the AI Landscape: GitHub Copilot and WSL Go Open Source!

In a groundbreaking move for developers and the open-source community, Microsoft has announced the open-sourcing of GitHub Copilot and the Windows Subsystem for Linux (WSL). This shift is especially remarkable for GitHub Copilot, which is now available under the permissive MIT license, allowing developers to inspect, modify, fork,

IBM Introduces Granite 3.0: High Performing AI Models

Listen to a podcast0:00/542.511021× Introduction On October 21, 2024, IBM announced the release of its most advanced family of AI models to date, Granite 3.0, at the annual TechXchange event. This third-generation flagship language model series is designed to outperform or match similarly sized models

Exploring Red Hat's Trusted Software Supply Chain: Addressing Challenges and Delivering Value

In today's rapidly evolving digital landscape, the security of the software supply chain has become a paramount concern for organizations worldwide. The increasing frequency and sophistication of cyber threats targeting foundational software components necessitate robust solutions that can safeguard the integrity and reliability of software delivery processes. Red

InstructLab Deep Dive Part 5

Index: InstructLab Deep Dive Part 1 What is InstructLab and what problem does it solve? InstructLab Deep Dive Part 2 Installing InstructLab on Fedora 40 with enabled CUDA. InstructLab Deep Dive Part 3 Installing InstructLab on Fedora 40 with Ansible InstructLab Deep Dive Part 4 Building InstructLab Podman/Docker container

InstructLab Deep Dive Part 4

Index: InstructLab Deep Dive Part 1 What is InstructLab and what problem does it solve? InstructLab Deep Dive Part 2 Installing InstructLab on Fedora 40 with enabled CUDA. InstructLab Deep Dive Part 3 Installing InstructLab on Fedora 40 with Ansible InstructLab Deep Dive Part 4 Building InstructLab Podman/Docker container

InstructLab Deep Dive Part 3

Index: InstructLab Deep Dive Part 1 What is InstructLab and what problem does it solve? InstructLab Deep Dive Part 2 Installing InstructLab on Fedora 40 with enabled CUDA. InstructLab Deep Dive Part 3 Installing InstructLab on Fedora 40 with Ansible InstructLab Deep Dive Part 4 Building InstructLab Podman/Docker container