Maina Wycliffe

Maina Wycliffe

@mainawycliffe.dev

Chief Builder at Garatropic. Building AI agents, skein-js, and practical tools. https://mainawycliffe.dev/

1.3K followers351 following560 posts

I am working on an exciting feature: local agents with skein-js. These will enable you to build and run your agents locally using Tauri or Electron, etc., eliminating the need for server management, all while utilising the same Agents Protocol. github.com/skein-js/ske... #typescript #langgraph

Angular 22.2 was released at the end of September 2026. Two additions stand out, both as developer previews: Error boundaries catch errors while rendering a section of a template and show a fallback view instead. Router Resources load route data with resources instead of resolvers, in parallel rathe

angulararchitects.ioAngular 22.2: Error Boundaries, Router Resources, and More - ANGULARarchitectsAngular 22.2 brings error boundaries with @boundary and Router Resources as developer previews. This article shows both using a demo and summarizes further changes from the changelog.

Agents aren’t just prompts + loops. Real‑world use means tools, decisions, persistence, human‑in‑the‑loop, and durability. Skein’s Building Blocks page shows how these primitives come together — with Postgres, Redis, and contracts instead of prose. skein-js.github.io/skein-js/building-blocks

Most discussions of continual learning in AI focus on one thing: updating model weights. But for AI agents, learning can happen at three distinct layers: the model, the harness, and the context. Understanding the difference changes how you think about building systems that improve over time.

share.googleContinual learning for AI agentsMost discussions of continual learning in AI focus on one thing: updating model weights. But for AI agents, learning can happen at three distinct layers: the model, the harness, and the context. Understanding the difference changes how you think about building systems that improve over time. The three main layers

One thing I've learnt building a production AI agent: persistent state isn't a nice-to-have — it's the foundation. Your agent sends a clarification message. The customer replies two days later. You need to resume exactly where you left off — not start over.

But persistence alone isn't enough. State gets stale. If something in their cart is now out of stock, say so. If you've restocked what they wanted, offer it. Durable context + real-time validation. You need both.