saved
The Multiplayer AI Manifesto
Hraness cites a source capture. The source author remains the source.
gist
Sergey Karayev's Multiplayer AI Manifesto argues workplace AI drifted into siloed single player chats that impose a context tax on every handoff, while laptop agents create org wide security risk. Citing HBS and Shopify evidence that team plus AI beats solo AI, it sets five principles: agents live beside shared work sessions, sessions stay open so practices spread, skills and benchmarks compound automatically, people stop routing answers agents already know, and cloud sessions resume so nothing starts from scratch. Enterprise plans and Slack bots are incomplete; cloud sandboxes, permission aware sharing, provider agnostic routing, and auditable governance are required.
ideas
- Siloed chats levy a context tax. Each private session-to-Slack-to-session handoff re-pays for context that a shared agent session would keep.
- True multiplayer is not settings sharing. Team plans that only sync prompts and integrations still trap learning in unindexed individual sessions.
- Agents belong beside the work, not in a side chat. Everyone on the work must talk to the same agent, with the same tools humans use, across surfaces.
- Open sessions compound; private ones stall. Shared chats, auto-skills, and tracked benchmarks turn one teammate's correction into team capability.
- Cloud, firewall, and audit are the hard constraints. Laptop agents stop when lids close and over-share credentials; multiplayer needs sandboxes, least-shared permissions, and year-later session provenance.
quotes
“Agents must live next to the work, instead of in their own chat or in Slack.”
“A human must never be asked a question that an agent already has the answer to.”
“When agent sessions live in the cloud, you can pick up every project exactly where you left off, with no context lost.”
“Every agent must be limited to the services you approve. That is only practical in locked-down cloud environments.”