9

Source document

[2026-01-28 06:57] lisamegawatts: Its currently failing to write a test script and fixing its emojis [2026-01-28 09:11] ajaxdavis: you got a diagram [2026-01-28 09:11] ajaxdavis: im making yet another style guide [2026-01-28 09:13] lisamegawatts: ok so i hate being a middle man between engineers and i just cut myself out and put claude and roo in a shared_doc.md to hash it out so they are going to work on how to interop [2026-01-28 09:13] lisamegawatts: they several proposals and a mermaid diagram going [2026-01-29 02:13] xenonfun: `ssh [email protected]` all your various SSH keys should work. (files: Screenshot_2026-01-28_at_9.12.56_PM.png) [2026-01-29 05:38] ajaxdavis: https://vercel.com/blog/agents-md-outperforms-skills-in-our-agent-evals [2026-01-29 08:10] lisamegawatts: i been saying! skills and tools are only good if invoked, but a role or mode is always invoked in a sense, it provides identity and that is context [2026-01-29 08:14] lisamegawatts: also the deep agents one is way cool because one better than a file is a file system. the claude agents.md is a role or mode, a job or title which is not the same as a person or agent, as people and agents can have many roles, and each time they get a new job they learn new skills. [2026-01-29 08:18] lisamegawatts: i will get off my soapbox, but you can read more about capabilities and continuity in my vrmai standards proposal 🤓 [2026-01-29 08:28] lisamegawatts: update [2026-01-29 08:28] lisamegawatts: (files: image.png) [2026-01-29 08:29] lisamegawatts: had to strip out most of the 3d as too hard to render, but progress! [2026-01-29 08:31] lisamegawatts: still trying to get swarm working on it, had some very fun experiments with roo and claude working out how to interop. looks like we are going with roo orchestrating swarm of clauds using the claude agent sdk, and they agreed on shared agent/mode definitions and shared skills and tool [2026-01-29 10:11] traves_theberge: 🤣 [2026-02-03 23:15] lisamegawatts: Q not sure where to put this but looked interesting https://x.com/alxfazio/status/2018744471857279438?s=46&t=OS71yaTltL7EclsAy81ZcQ [2026-02-03 23:15] lisamegawatts: <@187062569415081984> [2026-02-06 02:50] lisamegawatts: updates: self evolving bot has issues, new swarms are producing really good output (imho) [2026-02-06 02:51] lisamegawatts: research swarm on human team dynamics to be extrapolated into what not to do framework: https://app.affine.pro/workspace/092a7f99-7277-4456-b887-48e0ecab55dd/P1x-FtFHSDRd1fRS-wlbG [2026-02-06 02:52] lisamegawatts: Next research biological organizational strategies, then synthesize comparison [2026-02-06 02:52] lisamegawatts: end goal: cooperative agent framework, rules for how not to swarm [2026-02-06 02:59] lisamegawatts: https://app.affine.pro/workspace/092a7f99-7277-4456-b887-48e0ecab55dd/eKgKyY4RLdOtOqVEEbAV0 [2026-02-06 03:01] lisamegawatts: 47 page report over 200 sources and 6 key patterns identified