Tandem's future

[yyovil]

It seems like what we actually ever needed was a really really good MCP instead of spending time after building the same commodity experience that every terminal based AI Coding agent is trying to provide to the users. Also this is where actually these agentic systems differ otherwise they all are same as other. This also allows users to use their current subscription to extend their agent's capability if the MCP is built with the required competency.

[yyovil]

So i guess Tandem is going to get archived. I'll come back to this if i saw a better opportunity in building the agent from the ground up. at the moment, building the MCP is the best thing that i could do.

The goal is to prove that Penetration testing engagements is the kind of task that could be automated with the help of AI. regardless whether i build AI Agent or the MCP to do so.

[yyovil]

i'm really determined to take this step and move forward because the upside is that i could leverage the work done by dax and adam on OpenCode and then get to use those stealth models, Github Copilot pro plan (this will go down after Nov 1, 25) and i could also juggle in between qwen, github-cli and copilot cli as well. so no locking in mf. this way i'll get to spend more time on getting to know what the actual and unique problems there are to solve in this case.