Hello folks, I presented "High-quality IP Geofeeds using AI Coding Assistants and MCP" at the recent NANOG 96 <https://nanog.org/events/nanog-96/content/5683/> in San Francisco; YouTube video recording <https://youtu.be/x_vbTj_D91I?si=tTjqHQPthGIt9cnM&t=93>, deck attached as PDF. The "AI Skill" mentioned in the talk has been significantly improved since. Those interested in tuning their RFC 8805 IP Geofeed are encouraged to take it for a spin in a sandbox environment. Install the Skill in your favourite AI tool or CLI; then, ask the LLM to use the *'geofeed-tuner'* skill to tuner your IP geolocation CSV feed (paste from clipboard, provide a local file, or its public URL). [image: List operation and use of skills in Github Copilot CLI] Upon reading the SKILLS.md , the AI authors and runs Python 3 scripts locally to process the CSV and calls a geocoder MCP server to check the mentioned country and state/city names. It also detects bloopers such as duplicate inetnum entries pointing to different countries (see below). [image: Screenshow of Report with bloopers and geocoding improvemenents] The HTML report it generates has a "Tuning Mode" where you can ask it to look up the correct state/region codes that you are missing: [image: image.png] Happy tuning! And please send bouquets and brickbats to me, Sid Mathur ( sid@getfastah.com). I would love feedback on how to better fit your workflow or data visualization needs. This is experimental software in its v0.0.9; so a partial list limitations are as follows: 1. State/region code geocoding works well for 90% of the world's nations but has defects for the other 10% 2. The HTML web report in "Tuning Mode" needs improved usability and interaction design. Sincerely, Sid -- Siddharth 'Sid' Mathur Founder, Fastah Inc. The freshest IP data | API Quickstart <https://docs.getfastah.com/docs/quick-start> for AWS <https://aws.amazon.com/marketplace/pp/B084VR96P3> Text me: +1-617-500-7576 (M) Make an appointment: https://calendar.app.google/vkALeR3fB3bvscnx5