Writing
How to sandbox Claude Code with nono
Claude Code runs with your full user permissions. Every file, credential, and command is accessible. nono uses kernel-level sandboxing to enforce default-deny access with no escape hatch — setup takes 30 seconds.
Luke Hinds
Co-founder & CEO·
Why I built nono
I watched the same pattern play out with software supply chains. Now AI agents run with full user permissions and no boundaries. nono is kernel-level sandboxing that makes unauthorised operations structurally impossible.
Luke Hinds
Co-founder & CEO·
Why Your Classification Task Doesn't Need a $20M Model
Specialized classification models can outperform large general-purpose models at a fraction of the cost.
Luke Hinds
Co-founder & CEO·
Software Supply Chain Security in the Age of AI Agents
Exploring the challenges and solutions for ensuring software provenance with AI agents and the new paradigm they introduce.
Luke Hinds
Co-founder & CEO·
Local Language Models and security: Separating Fact from Fiction
Dispelling common myths about the security implications of running large language models locally versus in the cloud.
nolabs Team
nolabs·
Train a 4B Model to Beat Claude and Gemini at Tool Calling
A deep dive into deepfabric dataset generation, and how to train your own 4B model to outperform Claude and Gemini using DeepFabric.
nolabs Team
nolabs·
Everything you wanted to know about Tool / MCP Calling in Large Language Models
A comprehensive guide to LLM function / tool calling.
Luke Hinds
Co-founder & CEO·
Interpreting Learning Rate Values
What Numbers Like 5e-6 Actually Mean.
nolabs Team
nolabs·
Why System Prompts will never cut it as Guardrails
Explore the hidden challenges of system prompts in AI and how they can undermine your safety mechanisms.
nolabs Team
nolabs·
Announcing Always Further and our Pre-Seed Investment
Always Further raises $1.8M to build infrastructure for reliable, efficient AI agents.
nolabs Team
nolabs·
We need to talk about LLM's and non-determinism
LLMs are often called non-deterministic, but their neural networks are actually fixed mathematical functions. The randomness comes from sampling strategies like temperature and top-p, not the model itself.
nolabs Team
nolabs·
Just How Resilient Are Large Language Models?
Exploring how LLMs survive corrupted parameters, bit flips, and aggressive pruning — and what their fault tolerance reveals about the parallels between artificial and biological neural networks.
nolabs Team
nolabs·