I handle pipeline and partnerships at Eduba, and I'm asking the room for help on this one. We've got a shot at federal funding to run an AI capability-building pilot inside a public sector agency. Same playbook we've used with Pacific Life, Colgate, and KPMG UK, now pointed at the government workforce. The piece we're missing is the partner. State, county, or federal. An agency with legacy pain, a workforce ready to actually use AI the right way, and a leader willing to try something real. This is grant-funded. No procurement cycle, no sales pitch. The agency gets trained, equipped, and supported. We get to prove the model in public with the people who need it most. Timeline is tight. We need a partner locked in over the next few weeks. If you know someone in a state CIO office, a county agency, a federal program, or anywhere a civil servant is losing sleep over their tech stack, DM me. I'll personally work with you to figure out the best way in, whether that's a three-way intro, a short brief I can send you to forward, or a call with all of us on the line. And if the intro you make is the one that lands the partnership, you get lifetime VIP in the community on us. No cost, ever.

Quick question for the community: Jake emphasizes that file engineering is the foundation of powerful AI systems and I agree. But with new tools like Claude CoWorker and Claude Management AI for building agents, I’m a bit confused. These tools seem to simplify everything, but aren’t they just abstracting what strong file engineering already does (context structuring, memory, workflows)? How do you explain or demonstrate that solid file engineering can achieve similar (or even better) results than these tools? Curious to hear your thoughts 👇

When you move from building a static knowledge base to running an active, agentic workflow, something interesting happens. Jake's method proves once again you don't need bloated apps and proprietary solutions to get results. In this approach, you don’t need a specialized app to manage your AI. The folder system itself becomes the agent architecture and the user interface. Here’s how it plays out in practice. The Filesystem Is the Orchestrator In advanced agentic workflows, you are not just storing notes. You are running multi-step production pipelines — turning a script into animation, video creation pipelines, or handling complex sequences. Jake’s ICM handles this with the basic operating system filesystem. Numbered folders manage stage sequencing. Folder hierarchy controls context scoping so the AI only sees exactly what it needs. The output of one folder flows naturally into the next as input. Because the folder structure already organizes the logic, routing, and execution, adding Obsidian creates a redundant interface layer on top of a system that is already complete. As Jake puts it, what is simpler than a folder. VS Code Gives You a Leaner Setup For directing these workflows, many move to Visual Studio Code paired with Claude Code. VS Code shows you the raw folder tree and gives direct terminal access without extra layers. Obsidian is built as a personal knowledge management tool with a visual, plugin-heavy interface. When your setup is busy running Python scripts, scheduled checks, and processing files, a lightweight developer environment fits the factory floor better than a note-taking app. Plain Text Keeps Everything Clean A key principle here is plain text as the interface. Any tool that can read a text file should be able to participate without proprietary formats or hidden databases. Obsidian can alter how your files work by adding its own metadata and formatting requirements. Staying purely in local folders keeps the entire workflow transparent, untouched, and fully portable.

It happen to me, i used my 20$ usage token for the first time in 4days. I am questionning myself : how can i plug ollama to claude code until my usage are back ? Which model is the best ? And is it possible to use only Anthropic for few task and change llm automatically for others ? (Exemple : thinking, planning = anthropic , coding = a free llm )

I build this with my uncles. The most important things that you cant really see his the intent scoring system on the assistant Here is the link of the video https://www.linkedin.com/posts/eytan-levy-problemsolver_rag-lawyers-ai-ugcPost-7444783646265368576-g5xl?utm_source=share&utm_medium=member_android&rcm=ACoAABIn1z4BsesgFBocIqpi0fOr44pAN2Qxoyw