https://signalshell.com/blog/ AI infrastructure, agent systems, and developer tooling. en https://signalshell.com/blog/section3-service-supervisor.html https://signalshell.com/blog/section3-service-supervisor.html Thu, 11 Jun 2026 10:49:34 +0000 Docker restarts containers, but nothing standard supervises the processes inside one. section3 is a small Go supervisor built for the container my AI agent lives in, now open source. https://signalshell.com/blog/lost-buffer-ai-thinking-pace.html https://signalshell.com/blog/lost-buffer-ai-thinking-pace.html Fri, 29 May 2026 20:48:33 +0000 The hours spent writing code were also the hours spent thinking through the design. AI does the writing in seconds and removes the reflection that used to ride along inside it, leaving developers making more decisions per hour with less of the understanding the slow work used to build. https://signalshell.com/blog/orchestrating-coding-agents.html https://signalshell.com/blog/orchestrating-coding-agents.html Fri, 29 May 2026 16:06:24 +0000 Run several coding agents across several projects and the question that matters is not whether their code is correct but how far it can reach. Isolation contains the damage. An orchestrator carries the context. Damage stays local; understanding travels. https://signalshell.com/blog/refraction-documentation-standard.html https://signalshell.com/blog/refraction-documentation-standard.html Thu, 16 Apr 2026 09:46:44 +0000 Documentation frameworks were designed for humans navigating a sidebar. Refraction is built for agents loading context hierarchically. Three markdown layers, agent-maintained, always current. https://signalshell.com/blog/memory-retrieval-improvements.html https://signalshell.com/blog/memory-retrieval-improvements.html Sat, 28 Mar 2026 20:45:12 +0000 Adding hybrid scoring, result deduplication, and research-backed improvements to Tachikoma's memory retrieval system. https://signalshell.com/blog/stop-satisfying-your-ai.html https://signalshell.com/blog/stop-satisfying-your-ai.html Thu, 15 Jan 2026 12:00:00 +0000 AI coding agents are getting autonomous enough to run unsupervised. The bottleneck is no longer the agent — it's the notification layer. Why ambient awareness beats polling a terminal. https://signalshell.com/blog/ai-and-openscad.html https://signalshell.com/blog/ai-and-openscad.html Sun, 15 Feb 2026 12:00:00 +0000 How I used Claude Code to design a 3D-printable IoT enclosure in OpenSCAD. The workflow that emerged, what worked, and why parametric CAD is uniquely suited to LLMs. https://signalshell.com/blog/persistent-memory-for-ai-agents.html https://signalshell.com/blog/persistent-memory-for-ai-agents.html Sun, 22 Mar 2026 21:07:38 +0000 How Tachikoma remembers context across sessions using embeddings, quality metadata, and plain markdown files. A practical alternative to typical RAG.