Yes, I think that would be helpful. I see there were a lot of new additions over the last two days. I was wondering how it was possible for a single person to be doing all this work.
One thing that isn't "obvious" from the About This Project page is that LLMs were involved. Perhaps a pointer to your first post, or a summary in the page would at least let people know a bit more about the project.
I like the stated approach.
Approach
Every guide published here is:
- Verified on real hardware — No copy-paste from wikis. Every command is tested on actual MikroTik devices running RouterOS 7.x.
- Step-by-step — Written so you can follow along from start to finish without guessing.
- Consistent — Every guide follows the same structure: overview, prerequisites, instructions, verification, and troubleshooting.
I am curious how you leverage both Claude and ChatGPT. Do you have them check each other's work?
Even with the agentic loop, some hallucinations (or unverified info learned from internet sources, e.g. Community Tips) are still making it through. The example I gave in which book to buy - #15 by Buckeye, while the point of the section was that on a router with two switch chips, and having switch-ports belonging to the same vlan span the two switch chips will result is traffic being forwarded via the CPU is accurate, but the example routers included the RB5009, which is a single switch chip router (you can verify that in the MikroTik RB5009 block diagram here). Then this incorrect information is repeated in
I did a google search for "RB5009 mulitple switch chips" and wasn't able to find any reference to multiple chips in the RB5009, so it seems odd to me that the LLMs would have come up with that idea. But I have seen far worse hallucinations than that.
Did you have to come up with a high level outline? It is an interesting project.
What many don't like is users getting AI generated configs that have hallucinated commands, and then coming here to have people debug them.
How do you orchestrate the verification in your "home lab"? Are these simulations with EVEng or GNS3 or do you have real routers, and how is it automated? Can the LLMs get real time feedback from the simulations?