Stop Drowning in AI News

AI educator and agentic workflows specialist Cole Medin built an AI Content Engine, refined over more than six months, that monitors hundreds of sources including Reddit, YouTube, RSS feeds, Hacker News, GitHub repositories and changelogs to deliver a daily digest filtered to a user's own tech stack and priorities. The system is distributed as a skill workflow called build-signal-engine that has a coding agent interview the user for roughly 10-15 minutes to define positive and negative filtering rules, then scaffolds and tests an ingestion pipeline. Medin reported that processing 864 posts cost $0.03 for initial Jev calls and $0.42 for subsequent Claude Haiku interactions, with deterministic deduplication and rule-based culling applied before any model call.

The AI firehose has a signal problem AI’s relentless pace of releases—new models, GitHub repositories, research papers, and product updates—overwhelms even the most dedicated developers. Every week, multiple breakthroughs emerge, making it nearly impossible to keep up with what truly matters without drowning in the noise. Raw feeds and generic roundups only exacerbate the problem. They report everything new, creating either cognitive overload or critical blind spots. These aggregators lack the crucial context to filter for a specific individual’s tech stack, active projects, or content strategy, leaving developers to sift through mountains of irrelevant data. Cole Medin, an AI educator and agentic workflows specialist, tackled this directly. For over six months, he’s refined an AI content engine that monitors hundreds of sources—from Reddit and YouTube to Hacker News and GitHub changelogs—and distills them. This system's promise is a short, daily digest, meticulously shaped around the user’s own priorities. It doesn't attempt to summarize the entire AI internet; instead, it leverages a "signal engine" to identify and deliver only the high-value updates relevant to your specific workflows, saving several hours per week of manual filtering. Build a filter that knows your work Cole Medin, an AI educator and agentic workflows specialist, recognized this "signal problem" and built a solution: an AI Content Engine . Instead of offering a static open-source repository, Medin provides a unique skill workflow called build-signal-engine . This skill doesn't just hand you a system; it guides a coding agent to build one tailored to your needs. When you execute build-signal-engine , the coding agent "grills" you. It asks about your specific tech stack, active coding workflows, research interests, and content strategy. This detailed interview process defines explicit positive and negative filtering rules—what to follow, and crucially, what to exclude. This upfront investment is deliberate. You spend roughly 10-15 minutes answering focused questions, allowing the agent to scaffold and test an ingestion pipeline from scratch. The system then monitors hundreds of sources—Reddit, YouTube, RSS feeds, Hacker News, GitHub repositories, and changelogs—and delivers a concise daily digest. Medin’s engine, operating for over six months , reduces hours of manual triage each week. It’s a powerful example of how investing in a custom, context-aware filter can transform the overwhelming AI news flow into an actionable, high-signal stream. Cut noise before the models see it After a coding agent defines your custom filter, the next step is ruthlessly culling noise before expensive model calls. A processing order ensures efficiency: gather all sources into a central store, deduplicate previously seen items, then apply deterministic rules to remove stale or low-value posts. Only then do models engage. This two-stage triage begins with a fast, initial pass. An agent-powered system like Jev https://www.stork.ai/en/jev makes quick, obvious decisions, dropping clear fluff or flagging definite keepers. Items that fall into a grey area then move to a more capable language model, such as Claude Haiku https://www.stork.ai/en/claude-ai-workspaces , for nuanced relevance checks and digest preparation. This layered approach offers significant cost savings. For example, processing 864 posts cost Cole Medin just $0.03 for the initial Jev calls and $0.42 for subsequent Claude Haiku interactions. Stripping unnecessary text before any model call is crucial for these economics. Costs will always vary with volume and input length, but the principle holds: maximize deterministic filtering to minimize token usage. This system, part of Medin's broader second-brain setup, demonstrates how to make AI news consumption sustainable. For related insights, explore GitHub - coleam00/second-brain-starter https://github.com/coleam00/second-brain-starter . Enjoying this? Get one like it in your inbox each morning. one email a day · unsubscribe in two clicks · no third-party tracking Make the daily digest earn its place A configurable source list forms the backbone of your personalized digest, pulling from diverse origins. Ingestion routes can include APIs, command-line tools, MCP integrations, or services like Apify. You can aggregate: - RSS feeds - YouTube channels - Hacker News discussions - GitHub repositories - Research-paper feeds - Product changelogs Even a finely tuned personal engine demands ongoing care. Expect to dedicate time to tuning, source maintenance, and regular checks for missed stories or false positives. Keep the pipeline understandable; this simplifies diagnosing rate limits or broken integrations. To start, choose a few key sources relevant to your work. Document precisely what constitutes "useful" information and what counts as noise. Then, build a small version of this system in a fresh project using the build-signal-engine skill. Refine your rules based on the actual digest output, iterating until the signal is clear. Frequently Asked Questions What is an AI signal engine? It collects updates from chosen sources, filters them against your interests, and delivers a concise digest of the items worth your attention. How does Cole Medin’s skill build the system? The skill interviews you about sources, priorities, and filtering rules, then guides a coding agent through building and testing a customized pipeline. How does the engine reduce AI processing costs? It deduplicates items and applies simple code-based rules before using models, so fewer, shorter items reach more expensive reasoning steps. Can I use sources beyond RSS feeds? Yes. The workflow can connect to sources such as Reddit, YouTube, Hacker News, GitHub, and changelogs through APIs, tools, or scraping services.