Two hours, live, building a company brain from scratch with Austin Distel. Your AI does not have a capability problem. It has a context problem. A company brain is plain files in a folder your team already owns, read by an AI that knows how to route a question to the right one. The folder we handed out and the prompt that runs the setup are both on the same page here: https://os.agenticsociety.com/company-brain To connect with Austin directly: AgenticSociety.com/meet For those who want join the online Agentic Society membership to join our next 90 day sprint, details and links to pricing and how to join can be found here: https://agenticsociety.com/pricing If you have questions about joining this level, online or in-person in Austin, TX, feel free to send me a direct message here on Skool or email: [email protected] CHAPTERS 00:00 Welcome, and how the two of us got here 03:00 The thesis: a business that thinks for itself 05:00 Own your data, stay model-agnostic 07:00 Hire your AI the way you hire a person 10:00 The tooling floor, and what it costs 12:00 Why build a company brain: the six reasons 16:00 Rot and drift, the two ways AI memory fails 18:00 V1, V2, V3: you, your team, your agents 20:00 Security, HIPAA, access and governance 21:00 Shared Drive, not My Drive, and why 25:00 What to gather first 27:00 Inside the structure: CLAUDE.md, markdown, the inbox 29:00 Demo: the bookkeeper run 31:00 How the brain onboards itself 33:00 Permissions in practice 35:00 Employee onboarding as the design spec 36:00 The HR director hack 40:00 soul.md, CLAUDE.md, and a chief of staff named Riff 45:00 The daily close, and the mistakes log 48:00 Evergreen versus ever-growing context 53:00 Version control and change logs 56:00 Demo: asking a live brain for the vacation policy 1:03:00 Demo: the starter prompt and the harvest

Hi folks, I'm new to AICOS. I've been using a third party solution that does a lot of what AICOS does, but its getting expensive, so i'm looking at ways to offsite and outwrite replace it. The biggest thing is to identify items that should go into my todolist and create drafts for emails on demand. the problem is that is a web front end that makes it easier to move through todolists, and send emails directly after the draft is reviewed and edited. i'm wondering if anyone has done anything liek that yet with AICOS. What i've gotten so far is a Mark down todolist and markdons of email drafts that I have to then copy and paste into Outlook (in my case) to send.

Hey all - hope you’re enjoying the summer season and whatever festivities you enjoy. Been watching some FIFA World Cup here :) Question specifically for those who continue learning, growing and sharpening their AI skills with : - Claude Cowork Or - Claude Code Or - something else? - Answer the poll & Share why you make your choice. Totally personal opinion - Claude Cowork is awesome - and still I lately rarely use it because Claude Code is that much better. For me, it’s my AI Chief of Staff and runs / routes tasks to my 145+ AI agents as needed. Which is kinda crazy because Claude Cowork is also so incredibly powerful when you know how to use it - and you actually don’t really need to be technically at all to use it. Claude Code - it definitely helps to be more technical. BUT frankly once you get over the learning curve of connecting with a GitHub repository from the Claude Desktop and pair it with an AICOS Atlas Folder, and if you’re willing to learn a *little bit* about how to use it in Terminal, Obsidian, and other tools like Cursor… It makes Cowork look like child’s play. Which must then mean that regular Claude Chat is like an infant baby by comparison. Which is quite aligned with its relationship with Cowork. Cowork is like Claude Chat’s slightly older brother who figured out how to do a lot more and Chat is still in diapers always forgetting where it left its toys. AKA AI amnesia. Cowork is already figuring this out once you take the time to train it. Like you would a new hire. But (IMO) it’s still looking up to its bigger older smarter better brother Claude Code. Who’s can run laps around Cowork when it’s focused and not distracted building new shiny toys that it can sometimes make on its own or if left without proper direction. But - and this is a serious caveat - a lot of people do not really need to be using Claude Code. Especially when they are getting started. Just getting good and comfortable with Claude Cowork is often far more than enough to do what the majority of people need - obviously this depends heavily on the person or company goals.