I feel like we’re over-capitalizing a problem that could be solved with better protocols. If the "Git successor" is just a wrapper to help agents not hallucinate their own worktrees, it feels like a very expensive solution to a context-window management problem.
100%
MCPs truly give the agent tools and allow the agent to make better informed decisions given you can have configured the right MCP tools. Skills are good for knowledge and general guidelines. They give context to the agent, and I have seen some skills being excessively long that could into eat into the context window of the agent.
This tool https://protomcp.io/ helps a lot with testing MCP servers before integrating into the agent workflow. You can even see the agent call different tools in real time and view the trace.
What’s the expected cost-efficiency? With the current pricing gap between Sonnet and Opus, the biggest factor for adoption (if up for adoption) will be where Mythos lands on the price-per-token scale
Love the ingenuity here, but man, this feels like an easy way to get banned. Anthropic’s telemetry is going to see those atypical request patterns through a tunnel and probably flag it. Definitely cool for a side project, but I wouldn't risk my main Pro account on it.
Coolest variation of this: If the prisoners have a spy (like a friendly warden) who is allowed to enter the room first and swap the contents of just two boxes, the prisoners' survival probability jumps from 31% to 100%.
By swapping two items, the spy can break any cycle longer than 50 into two smaller cycles. Since every prisoner is then guaranteed to be on a loop of length 50 or less, everyone is mathematically certain to find their number.