Smarthome & Home Lab

The Bookmark Everything app. Hoard links, notes, and images and they will get automatically tagged AI.

Thanks to the Raspberry Pi, we have easy access to extremely inexpensive machines running Linux that have all kinds of GPIO as well as various networking protocols. And as the platform has improved…

When I started this project, I knew deep in my heart that we could get an LLM plus tool-usage to solve arbitrarily complex workflows. I still believe this is possible, but I’m no longer convinced this is actually a good solution. Some problems are just vastly simpler, cheaper, and faster to solve with software. This post talks about our approach to supporting both code and LLM-driven workflows, and why we decided it was necessary.

Most of the extensions to our internal agent have been the direct result of running into a problem that I couldn’t elegantly solve within our current framework. Evals, compaction, large-file handling all fit into that category. Subagents, allowing an agent to initiate other agents, are in a different category: I’ve frequently thought that we needed subagents, and then always found an alternative that felt more natural. Eventually, I decided to implement them anyway, because it seemed like an interesting problem to reason through. Eventually I would need them… right? (Aside: I did, indeed, eventually use subagents to support code-driven workflows invoking LLMs.)

Understanding large GitHub repositories can be time-consuming. Code-Analyser tackles this problem by using an agentic, approach to parse and analyze codebases.

Playing a video game online is almost second nature now. So much so that almost all multiplayer video games have ditched their split-screen multiplayer modes because they assume you’d rather …

Although my model of choice for most internal workflows remains ChatGPT 4.1 for its predictable speed and high-adherence to instructions, even its 1,047,576-token context window can run out of space. When you run out of space in the context window, your agent either needs to give up, or it needs to compact that large context window into a smaller one. Here are our notes on implementing compaction. This is part of the Building an internal agent series.

#110: Understanding Model Context Protocol

AWS is changing the cost equation for enterprises weighing up where to put AI infrastructure for processing sensitive data.

The era of fragmented plugins is ending. Model Context Protocol is changing the game, offering cross-model compatibility, richer context and lower overhead.

Zero cost. Zero data sharing. Full privacy.

Using your Claude Max/Pro subscription as an API Key

Appwrite is an open-source platform for building applications at any scale, using your preferred programming languages and tools.

Smarter retrieval strategies that outperform dense graphs — with hybrid pipelines and lower cost

Agentic coding is incredibly powerful, but knowing when to step in and take charge is key

The biggest gap between AI agents and human intelligence is the ability to learn. There are various emerging approaches to support continual learning for AI ...

The best defense against prompt injection and other AI attacks is to do some basic engineering, test more, and not rely on AI to protect you.

Realtime collaboration and sharing, on the web and in Obsidian.

Explore data with Python & SQL, work together with your team, and share insights that lead to action — all in one place with Deepnote.

Take Control of Your Research, Privately

Skip to Translations Introduction  Generative AI tools (GenAI)  that allow users to rapidly generate new and creatively unique media (video, sound, text, and image) are increasingly being use...

Generate structured system designs using AI. Describe your requirements and get detailed architecture diagrams, components, and tech stacks powered by Gemini and Langchain.

GitLab.com

EVE-NG Professional Edition Latest Now: Release 6.4.0-13, 1 October, 2025   EVE-NG PRO platform is ready for today’s IT-world requirements. It allows enterprises, e-learning providers/centers, individuals and group collaborators to create virtual proof of concepts, …

Developers welcome AI for toil and boilerplate but demand human oversight for high-stakes decisions and resist automation in mentoring and strategic work

This post is a summary of my QConSF talk and some thoughts on what could come next, refined in a discussion with ChatGPT after getting it…