Need Help & Advice !!
retained text
(used chatgpt to structure my ask only) Hey guys! I am tired of keeping 10 tabs open just to keep up with AI. So I’m building an AI Agent for it. Every day, there’s something new a model release, research paper, startup, funding round, founder announcement, open-source project, or some shift in the tech ecosystem. The problem is that the information is scattered across X, LinkedIn, Hacker News, research papers, company blogs, GitHub, news sites, etc. So I’m thinking of building an **AI-powered daily intelligence agent** that: → Collects the latest and most impactful AI/ML + tech news → Tracks companies, founders, research papers, models and open-source projects → Filters out repetitive / low-value information → Evaluates how relevant and impactful each story actually is → Summarizes the important stuff instead of just dumping links → Sends me a personalized digest every morning at **8 AM** I don’t want this to be another “summarize the internet” LLM wrapper. I’m specifically thinking about building it with **proper guardrails and an evaluation framework** measuring things like relevance, factuality, source quality, duplication, and whether the final summary actually captures *why something matters*. I’m also exploring whether **Jev** could be useful for evaluating/ranking the relevance of collected information. For the product itself, I’m thinking about two parts: **1. A web app** Where you can see the latest stories, trends and sources in near real-time, potentially using RSS + other data sources. **2. An automated email digest** A personalized, curated briefing delivered every morning. The architecture is still very much in the **“thinking and experimenting”** stage, which is actually why I’m posting this. I’m considering making the whole thing **open source** and eventually building a community around it where people can contribute sources, evaluation criteria, agents, integrations, and ideas. If you’ve built something similar, worked on **AI agents, news/research aggregation, RAG, evaluation systems, RSS pipelines, or agent guardrails**, I’d genuinely love to hear how you approached the architecture. **If you want to build this with me, have ideas, or just think I’m approaching it completely wrong — drop a comment or connect with me.**
document fetched 2026-09-26 00:18 · canonical https://reddit.com/r/AI_Agents/comments/1wpp1xh/need_help_advice