Run a 24/7 local agent that curates news digests
The author runs Hermes Agent on a Proxmox VM with a local Qwen model, monitoring RSS feeds, curating news digests, and reporting via Telegram.
Done withHermes Agent
- What they did
- Hermes Agent runs on a small Proxmox VM (4 vCPU, 8 GB RAM, no GPU) and calls a Qwen 3.6 MoE llama-server on the author's RTX 5060 Ti. The author sent one Telegram message asking for a daily 8 AM cron job. Hermes created it, pulled RSS and MCP data, compared against its memory, and sent briefings to Telegram topics.
- How it went
- The agent delivers daily briefings and a 9 AM startup/VC digest at no API cost. Memory builds up over weeks into trend-aware reports, and Hermes writes reusable skills. The post gives no formal quality measurements.
- Worth knowing
- Don't over-prompt the cron job: a clear goal worked better than detailed steps. The fast MoE model beat the dense 27B, which was about 3x slower.
Try it yourself with Hermes Agent
Set up a news digest for me. Monitor these RSS feeds: [list of feed URLs], and every [morning at 7am / every 4 hours] pick the [5-10] most relevant stories on [topics I care about], skipping duplicates and clickbait. Send me a short digest on [Telegram / other channel] with a one-line summary and link for each story, and confirm by sending me a test digest now.
Want help getting this working? Join the Vibe Coding Academy ↗ Run by Alex Finn, who also runs this site.
2 more accounts of this
Source: njannasch.dev · Undated
Related use cases
Set up a home chat server and manage accounts from your phone
A Hermes agent running on a local Qwen model set up an Open WebUI server and handled account management via phone, giving the author's teen their own account.
Flag documentation drift from merged PRs weekly
A weekly Hermes cron job checks merged PRs for changes to tool schemas, CLI commands, config options or env variables to flag docs needing updates.
Run a weekly competitive-monitoring briefing for product management
A product manager had Hermes produce a weekly competitor briefing; its skill rewrote itself four times and by week six surfaced patterns he had missed manually.
Find connections across 100 non-fiction books
Claude Code browses a personal library with CLI tools and surfaces trails of related excerpts across books.
Try Devin on research tasks (mostly failed)
Answer.AI had Devin fetch an author's recent papers and research transcript summarization; it got stuck parsing HTML and returned generic results, rated a failure.
Run web research jobs and capture screenshots into a doc
A reviewer tested Operator on analyzing web-design posts, summarizing AI news, listing website builders with prices and screenshotting sites into a Google Doc, finding login and CAPTCHA handoffs frustrating.
Get 5 like this every morning
The best things people got an AI agent to do, each with the prompt to try it.