{"slug": "i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i", "title": "I built a self-hosted AI agent with a 30-min self-improvement loop — here's what I learned", "summary": "A developer released Hyper Nexus, a self-hosted AI agent built with FastAPI, SQLite, PyTorch, and a vanilla JS WebUI, designed to improve through a 30-minute self-improvement loop. The project, which comprises roughly 60,000 lines of Python code, prioritizes tool execution and state management over the LLM call itself. The developer acknowledged shipping version 0.1.0 without automated tests and is now working on an evaluation harness to measure the agent's actual improvement.", "body_md": "Six months ago I started building an AI agent I actually wanted to use.\n\nNot another LangChain wrapper — a single, self-hosted system that gets\n\nmeasurably better the more I work with it.\n\nThis week I cut the v0.1.0 release.\n\nWhat it is\n\n[Hyper Nexus](https://github.com/Hsosn/HYPER_NEXUS) is a self-hosted AI\n\nagent with:\n\n`pip install`\n\n.Stack: FastAPI, SQLite, PyTorch, vanilla JS WebUI. ~60K LoC of Python.\n\nWhy I built it\n\nWhen I started this, I thought: why not try to model something close\n\nto how humans actually think? The result isn't fully polished, and\n\nthere are real shortcomings — but I'd love feedback so I can keep\n\nimproving it. This is going to be an open-source project, and I want\n\nit to grow with the people who use it.\n\nWhat I learned building it\n\nLesson 1: The hard part is not the LLM call.** It's everything around\n\nit — tool execution, error recovery, state management, the agent's\n\n\"short-term memory\" of what it's already tried, the user's long-term\n\ncontext. The actual prompt is maybe 5% of the code.\n\nLesson 2: Tests matter even for solo projects.** I shipped v0.1.0\n\nwith zero automated tests. I regret this. If you're reading this and\n\nconsidering the same — don't.\n\nLesson 3: Don't promise self-improvement you can't measure.** I have\n\na 30-min heartbeat that does *something*. Whether it actually makes\n\nthe agent better at your task is unmeasured. I'm working on an eval\n\nharness to find out.\n\nWhat's next\n\nIf you try it, please open an issue — that's the only way I can\n\nprioritise what actually breaks vs what I *think* breaks.\n\nLet's make something meaningful.\n\nGitHub: [https://github.com/Hsosn/HYPER_NEXUS](https://github.com/Hsosn/HYPER_NEXUS)\n\nMIT licensed. PRs welcome.", "url": "https://wpnews.pro/news/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i", "canonical_source": "https://dev.to/smith_falcao_5faa19464e7f/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-heres-what-i-learned-3l5p", "published_at": "2026-06-03 17:12:09+00:00", "updated_at": "2026-06-03 17:42:37.904943+00:00", "lang": "en", "topics": ["ai-agents", "artificial-intelligence", "ai-tools", "ai-products", "ai-research"], "entities": ["Hyper Nexus", "FastAPI", "SQLite", "PyTorch", "LangChain"], "alternates": {"html": "https://wpnews.pro/news/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i", "markdown": "https://wpnews.pro/news/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i.md", "text": "https://wpnews.pro/news/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i.txt", "jsonld": "https://wpnews.pro/news/i-built-a-self-hosted-ai-agent-with-a-30-min-self-improvement-loop-here-s-what-i.jsonld"}}