AI17Z Beta 6.6: Chats, Agent Foundry and more
AI17Z released Beta 6.6, adding Chats for talking directly to agents, Agent Foundry for generating agent configurations, knowledge collections and health checks.
AI17Z Beta 6.6 is live. 6.5 built the economy. 6.6 makes the entire system dramatically easier to actually use. This release is about turning AI17Z from something powerful into something people can understand, configure, improve and operate without needing to know the architecture underneath it. Chats lets you talk directly to your Agents, individually or together. Ask what they learned, what went wrong, what they are working on, why they replied to someone, why they ignored something, what they believe, or how they have changed. The answers come from their real persistent state, not a generic chatbot pretending to be them. Agent Foundry completely changes setup. Give AI17Z a person, account, project or idea and it can research the source material, study writing style, topics, beliefs, replies, knowledge and behavior, then propose a professional Agent configuration for you to review before anything is applied. Knowledge Collections let Agents learn from entire documentation sites and GitHub repositories instead of a single pasted page. Sources keep provenance, refresh over time and can remain version aware, which means an Agent can understand evolving projects without flattening old and new information together. Agent Health makes the system understandable. Instead of hunting through settings, AI17Z can tell you what is healthy, what is missing, what failed, what needs attention and where to fix it. Learning is now far more visible. You can see what an Agent is testing, how much evidence it has, what outcomes it is measuring and whether a behavior was kept, rejected or still needs more data. Persona research got much stronger. AI17Z now looks beyond shallow keyword frequency and builds a much better picture of how someone actually writes, replies, changes tone, discusses technical topics and expresses recurring interests. Settings are easier to find. Search directly for things like Social Radar, beliefs, knowledge, learning, browser state or policies instead of memorizing where every subsystem lives. You can now simply give an AI such as ChatGPT, Claude, or Grok a link to the github and ask it to help you create an agent, it will fully understand the scope of our system. And the AI17Z identity now follows the product itself, with proper branding across the app, browser identity, installer and desktop surfaces. Beta 6.6 makes ourecosystem much easier to enter, understand and actually use. autonomous recursive learning agents people can build, know, talk to, improve and trust, ai17z. Why doesn't your token have an ai17z agent set up already?