Nathan BrakeBlog
When work is a chat box, the desk becomes optional
How my engineering habits shifted to the phone, and the Agent of Empires release that takes the web dashboard out of beta.
The Hardest Part of Running a Small Business in the Trades
The admin work that piles up on tradespeople, and Clawbolt, an assistant built to take it on.
mozilla.aiAI Generated Code Isn’t Cheating: OSS Needs to Talk About It
Open source projects should be open about AI-written code instead of treating it as a secret.
mozilla.aiHow to deploy ChatGPT on your own hardware, the hard way
How I stitched together LM Studio, OpenWebUI, Cloudflare, and Google OAuth to run "BrakeChat" entirely on my own hardware.
Control LLM Spend and Access with any-llm-gateway
Budgets, analytics and access management for every LLM provider your team uses.
mozilla.aiRun Any LLM with a Single API: Introducing any-llm v1.0
A stable release of one interface for cloud and local models, with standardized reasoning output.
mozilla.aillamafile Returns
Mozilla.ai adopts llamafile to push open, local, privacy-first AI forward.
mozilla.aiStandardized Reasoning Content: A first look at using OpenAI’s gpt-oss on multiple providers using any-llm
How providers return reasoning output from gpt-oss, and how any-llm makes it consistent.
mozilla.aiIntroducing any-llm: A unified API to access any LLM provider
A Python library that talks to any LLM provider through one interface.
mozilla.aiIntroducing Any-Agent: An abstraction layer between your code and the many agentic frameworks
Decouple your code from any one agent framework, and compare them on equal terms.
mozilla.aiBenchmarking DeepSeek R1 Models using Lumigator: A Practical Evaluation for Zero-Shot Clinical Summarization
Finding the smallest DeepSeek R1 model that still writes good clinical summaries.
mozilla.aiRunning an open-source LLM in 2025
The open-source LLM landscape, and how to evaluate a model before you rely on it.
mozilla.aiEnd of Sequence Token Explained
How is the pad token handled in training a transformer and what's the impact of setting the pad token to be the same as the eos token?