How Machinecraft Built a 36-Agent Company Memory

AI Engineer9m 58s
0 comments · 0 votesOpen discussionClose discussion
Sign in to join the discussion

    Video summary

    Rushabh Doshi describes a manufacturing company whose most important knowledge lived in a few people's memories. Staff turnover and informal handoffs made that dependence risky, so Machinecraft began capturing the company's history, decisions and customer context in a shared system.

    The resulting architecture uses 36 role-specific agents connected to common memory. A recurring consolidation cycle reviews new information, reconciles conflicts and preserves useful lessons so later agents can act with more context than a static document repository would provide.

    Doshi says the expensive part was teaching the system the company's actual operating knowledge, not model training or compute. Machinecraft has published a blank version of the architecture, but every organization must supply and govern its own truth. Product promotion is omitted.

    Original YouTube thumbnailWatch on YouTube

    Share this page

    Rushabh Doshi beside the headline 36 Agents Share Memory Framed in blue with WWW.ARTIFICIAL-INTELLIGENCE.VIDEO, 27 July 2026 and duration 9m 58s.

    Rushabh Doshi explains how Machinecraft turned scattered institutional knowledge into a 36-agent system with shared memory, role-specific workers and recurring consolidation.