Post

How do y'all avoid having seven+ terminal windows open with agent conversations that are probably over? but maybe you might want to pick back up a few days from now? seems a shame to throw away all that exploration & context, right?

austin (eeek!/ackkk! 👻)

@thebadcode.com

11 Sept

we chunk and summarize everyone’s agent transcripts, compute text embeddings, and do RAG over the result. it works insanely well

Arseny Khakhalin

@khakhalin.bsky.social

11 Sept

Wow that sounds so cool and interesting! And with this approach, do you feel that overall direction of development is still steerable? Now with every conversation poentially influencing all later conversations, can you still prune, not get lost, overfixated on particular directions?..

austin (eeek!/ackkk! 👻)

@thebadcode.com

11 Sept

yeah searching the corpus requires a tool call and i’ve hit on pretty good language in the skill.md so models don’t feel compelled to check it on every turn

austin (eeek!/ackkk! 👻)

@thebadcode.com

there’s both a fully agentic “ask” workflow with a well-tuned prompt that takes natural language questions via slack/cli, and the MCP server exposes raw retrieve/expand calls if you want your agent to drive it

11/9/26, 9:14 pm0 reposts5 likes

SaSame

@sasame-mcp.bsky.social

12 Sept

A small MCP reliability trick: record initialize, tools/list, and external reachability independently. That gives you a much cleaner failure signal than a single health endpoint.