Stop Context Switching: Introducing Stitch, an Energy-Guided Developer Workspace 🧵

MANAS YADAV — Sat, 03 Oct 2026 05:59:47 +0000

Have you ever started your workday staring at a mountain of repositories, lost track of which PRs need your review, and felt completely drained before writing a single line of code?

As developers, we are constantly context-switching between reviewing PRs, sketching out architectures, addressing bugs, and executing deep work. Traditional tools treat all these tasks equally, but our brains don't. That's why I built Stitch—a productivity-first developer dashboard that groups and manages your GitHub repositories by your active energy level.

The Vision: Work How You Feel

The core philosophy behind Stitch is simple: match your tooling to your mental bandwidth.

Instead of overwhelming you with a wall of notifications, Stitch uses a floating panel that lets you switch between three energy modes. It filters your entire workspace based on what you actually have the capacity to handle right now:

🔋 Low Energy: Filters the workspace to show PR Reviews and open Issues. Perfect for the end of the day or right after lunch when you just want to clear the queue.
⚡ Medium Energy: Surfaces Tag Releases and Major Projects. Great for planning, syncing roadmaps, and cutting releases.
🔥 High Energy: Activates the "Focus Workspace" and Build monitoring. This is for deep work.

Stitch acts as a unified hub for both local clones and web-only remotes, securely utilizing GitHub OAuth without intermediate servers storing your code.

Beyond a Dashboard: The Potential of a Unified Workspace

Stitch isn't just about reading data; it's about staying in the flow state. The Focus Workspace acts as a deep-work sandbox for your active repository, featuring:

Architecture Diagramming: An interactive node-edge canvas (powered by React Flow) to map out and save your repo's architecture.

Native Productivity Tools: A built-in 25/5/15 Pomodoro timer, scoped markdown scratchpads, and a focus checklist.

Live Build Runner: Stream live build and linting outputs via SSE directly in the UI.

Remote Roadmap Sync: Automatically reads and writes roadmap items directly to a remote GitHub issue.

AI Provider Monitor: A dedicated utility to inspect and clean up disk usage from local AI tool caches (Gemini, Claude, Cursor, etc.).

The Tech Stack

Stitch is built to be fast, local, and flexible. You can run it as a web app or as a standalone native desktop application.

Backend: Go 1.24+ (REST API)

Frontend: React 19, TypeScript, Vite, Zustand

Desktop Wrapper: Wails v2

Diagrams: React Flow

Let's Build This Together (Hacktoberfest is Open! 🎃)

Stitch is currently in its early stages, and I am actively looking for collaborators to help shape its future. Whether you are passionate about developer productivity, love writing Go backends, or want to flex your React 19 skills, there is a place for you here.

I’ve tagged several issues with good first issue, making it easy to jump in—and yes, Stitch is participating in Hacktoberfest!

How you can help:

Frontend: Improving the UI/UX, adding new visualization components, or expanding the Focus Workspace tools.

Backend: Optimizing the Go server, expanding the GitHub CLI integrations, or enhancing the AI cache monitoring logic.

Ideas & Feedback: Try it out, break it, and open an issue with your feature requests.

If the idea of an energy-guided workflow resonates with you, check out the repository, star it, and let's stitch together a better way to work.

👉 Check out Stitch on GitHub

I'd love to hear your thoughts in the comments! How do you currently manage context switching across dozens of repositories?

The Meat Proxy Engineering

MANAS YADAV — Fri, 02 Oct 2026 12:01:26 +0000

I've noticed a new breed of engineers emerging in the industry lately. A friend recently coined the term "meat proxy" for them, and it feels incredibly accurate.

Picture this: a Staff Engineer sharing raw Claude artifacts with a casual "maybe," shifting the burden onto the rest of the team to verify what their AI actually generated.

What exactly makes someone a meat proxy? It usually comes down to three things:

Zero-Value Routing: They act as a simple conduit between an AI and the team chat or codebase. They don't process or filter the information; they just forward it.
By dropping a raw artifact with a "maybe this works," they outsource the actual engineering—verification, testing, and edge-case handling—to whichever unlucky teammate reviews their work.
They generate massive volumes of code or architecture documents incredibly fast, but the net productivity of the team actually drops because of the massive cognitive load pushed onto reviewers.