In the first StyloMail article, I used Jev to ask specific questions about email and get typed, probabilistic answers back. Nimble, from Bespoke Labs, brings that technique to a model I can host myself through Ollama. From C#, I can send the same kind of state…

Yes, I'm getting on the Jev bandwagon. I wanted to see how far I could get using it to build a behavioural spam-blocking proxy in .NET Core. Something that looks at mail in both directions: what arrives in your inbox, and what leaves an account that might have…

Signal Shingle: a novel architecture for high-performance ASP.NET multi-widget sites

Part of the Stylo.Bot release series. Behaviour-Aware ASP.NET UI covers the server-rendered dashboard surface; Finding and Fixing Unbounded Growth covers the reliability discipline behind bounded state; Learning to Get Faster covers the earlier adaptive-cache…

The StyloAgent Workflow: How I Build and Manage Large (and HUGE) Systems Using Code LLMs (Part 2)

Part 1 was the shape of the thing. This part is what it does. A model only earns its keep under real work : dropping the fleet on a codebase it has never seen, pushing one change across repositories, and keeping a decision trail that survives a restart. This…

The StyloAgent Workflow: How I Build and Manage Large (and HUGE) Systems Using Code LLMs (Part 1)

This is Part 1 of two : the problem and the StyloAgent model. Part 2 : The Workflow in Practice applies it. Code generation stopped being the problem a while ago. The models write good code. They follow the conventions in the file they are editing, and if you…

Behavioural Inference: How I Learned to Stop Worrying and Love Probabilistic Systems

Most software still assumes the world will hand it clean inputs and stable rules. Production usually does neither. This post is the through-line behind a bunch of things I've been building: DiSE, Constrained Fuzziness, CFMoM, Reduced RAG, StyloFlow, The Ten…

MCP Is a Transport, Not an Architecture

"How do I build an MCP server?" is the wrong question. SDKs exist. An LLM can scaffold one in seconds. The real question is: What role should an MCP server play inside a system that must not lie? The scarce skill now is designing the system around MCP:…

Why LLMs Fail as Sensors (and What Brains Get Right)

LLMs (Large Language Models) are being used as sensors. That is a category error: using a probabilistic synthesizer where a deterministic boundary device is required. This isn't developers' fault. Most mainstream examples and tutorials lead with the simplest…

VideoSummarizer: Reduced RAG for Video (Shots → Scenes → Evidence)

Status: In development as part of lucidRAG. Source: github.com/scottgal/lucidrag Where this fits: VideoSummarizer is the orchestrator of the lucidRAG family, combining five pipelines into a unified video analysis engine: DocSummarizer - Documents (entity…

AudioSummarizer: Constrained Fuzzy Forensic Audio Characterization

Status: AudioSummarizer.Core is currently in development as part of lucidRAG, a forthcoming mostlylucid product for multi-modal RAG. The implementation is complete and working—this article documents the architecture and design. CLI integration will be…