Meet Jarvis,
My Vibe Manager

The system isn't one tool. It's three layers that compound — each one solving a different problem, together creating something that feels like having a project manager who actually gets it.

A docs/context/ folder in your codebase with markdown files: strategy, OKRs, bottlenecks, content plans, positioning.

This isn't documentation. It's the memory your AI assistant reads before every conversation. When you say "plan next week's content," it already knows your quarterly theme, your target audience, your upcoming events, and the bottleneck you're trying to break. No re-explaining. Ever.

A code editor that doubles as your AI operating environment. Point it at context files, access MCP servers (your database, Stripe, Linear, Notion), write code when you need it, draft newsletters to specific audience segments — things you would never have had time for before.

I use Wispr Flow for voice input — talking instead of typing. Brain dump everything on your mind, and the AI structures it into actionable issues. It's faster and more natural than any keyboard workflow.

And it compounds. Every correction you make becomes a rule or skill the AI remembers. Every friction point you encounter gets smoothed out. The system gets smarter about you over time.

The visual execution board your codebase can't provide. Linear gives you the Kanban view — Ideas, Planning, In Progress, Done. The stages that let you see at a glance where everything stands.

The magic: through Linear's MCP, your AI can read projects, create issues, set priorities, estimate effort, organize sprints — all through natural conversation. You talk. The board updates. That's vibe management.

"Therapeutic" isn't a word you expect about project management. But it's the most honest word I have for what this feels like.

As a founder, my stress level is usually high. There's always more to do than time allows. Ideas scatter across tools, notes, and mental to-do lists. You go to bed wondering what you forgot. You wake up already behind.

Vibe management changed that. Not by reducing the work — but by giving me a weekly rhythm that made the work feel manageable.

The sprint cycle is the heartbeat. Every Monday, I plan the week — brain dump everything, let the AI create and estimate issues, commit to what fits. During the week, issues move across the board: Backlog to In Progress to Done. Every day I can see exactly where I stand.

That progression is what makes it therapeutic. Not just a to-do list that grows — a board that visibly moves. You finish something, it moves. You start the next thing, it moves. By Friday, you can look back and see a week of real progress — not a vague feeling of "I was busy," but actual evidence of what got done.

This is the question I get most — and the one I take most seriously.

If your AI project manager agrees with everything you say, you've just automated confirmation bias. That's not a Jarvis — that's a mirror.

I'm fully aware of the sycophancy risk. I still feel 100% responsible for every outcome. But I've learned something important: if you ask the right questions, you can get very meaningful pushback.

I asked the AI to compare what I actually worked on in January and February against the OKRs I set at the beginning of Q1 2026. The context files had everything — my stated goals, my planned priorities, and through Linear, exactly what I actually spent time on.

It didn't catch everything. But it pushed back on important things:

None of this was prompted by "tell me what's wrong." It came from a specific question: "Where am I kidding myself about my Q1 priorities?"

That's the trick. "Does this plan make sense?" will always get a yes. Asking the AI to compare your stated goals against your actual behavior — with the data right there in Linear and the context files — gets something useful.

The AI is the advisor. You are the captain. You decide whether to adjust the plan, correct your behavior, or accept the current path. But at least now you're making that decision with data, not delusion.

I believe the workflow I'm describing here isn't just a personal productivity hack. It's a preview of how we'll all co-work with AI.

The pattern is simple: you break a project down, decide the direction, and then AI agents execute the tasks. At critical points where a human needs to be in the loop, they present you with a solution, give you options, and continue from your decision.

This has already started. AI agents are becoming users in the same tools we use. You can assign any Linear issue directly to Cursor, or mention it in a comment — and a cloud agent gets to work. It reads the context, pulls in the relevant details, implements a solution, and opens a pull request. It can work across multiple issues simultaneously, all before you've even opened the task yourself.

That's the shift. Agents aren't chat windows you paste problems into. They're teammates. They show up in your project board, your assignee dropdown, your comment thread — right next to you.

Think about that. You plan a sprint on Monday morning. By Monday afternoon, agents have already started working through the backlog. You review their solutions, accept or redirect, and move on. The tasks that need your judgment get your full attention. Everything else just gets done.

The captain sets the course. The crew executes. You intervene where it matters.

We're not there fully yet. But the gap between planning and execution is collapsing fast. What used to take a team of five and weeks of coordination is becoming one founder with context files, an AI that remembers everything, and a board full of agents ready to work.

The founders who set up this system now — who learn to organize context for AI, who build the habit of structured sprints, who develop the skill of directing agents — will have an enormous advantage as these tools get better. And they're getting better every month.

Cheers,
Ben