SteeplyOpinionated: A PR Review Tool for Tea Recipes (Returns 418, Obviously)

Вера Махлова — Sun, 12 Apr 2026 16:02:17 +0000

This is a submission for the DEV April Fools Challenge

What I Built

A pull request review tool that does absolutely nothing useful.
You paste your tea recipe. A senior engineer reviews it. Every step has critical issues. The PR can never be merged. It returns HTTP 418 — I'm a teapot.
Because someone had to build this. And that someone was me.

Demo

Code

VeraStorozhenko / steeply-opinionated

How I Built It

Built with Python + FastAPI because enterprise tea deserves enterprise tooling.
The reviewer is a highly sophisticated, enterprise-grade AI system — which is actually just a carefully curated list of passive-aggressive comments. No AI was harmed in the making of this tool. No AI was used either. Just pure, hand-crafted pedantry.
Sample review comments:

"Magic number detected. Extract 3 minutes to STEEP_DURATION_SECONDS environment variable."
"No error handling. What if the kettle fails? Where is the retry logic?"
"Side effects detected. This step mutates global state (the cup). Please isolate."
"No unit tests found for this step. How do we know boiling actually works?"

Every merge attempt returns HTTP 418 — I'm a teapot as nature intended.

Prize Category

Best Ode to Larry Masinter — because RFC 2324 deserves a production-grade implementation. Thank you Larry for your vision. This is the tool he always dreamed of. Probably.

Freelance OS Copilot: A Notion MCP-Powered Operations System for Freelancers

Вера Махлова — Mon, 23 Mar 2026 13:31:02 +0000

This is a submission for the Notion MCP Challenge

Freelance OS Copilot: A Notion MCP-Powered Operations System for Freelancers

What I Built

I built Freelance OS Copilot, a product that turns a freelancer's Notion workspace into an AI operations system.

Instead of treating Notion like passive storage, the app helps answer practical questions like:

What needs attention today?
Which project is slipping?
What should I tell the client?
What should I do next?

The product combines:

a dashboard with urgent and at-risk project signals
copilot quick actions for daily priorities and weekly reviews
write-back actions for project creation, logging hours, and saving follow-up notes
Project Rescue Mode, the signature workflow

Project Rescue Mode is the standout feature. When a project is risky, the app does not just flag it visually. It:

explains why the project is risky
proposes an internal recovery plan
drafts a client-safe update
highlights the immediate next move

That turns the app from “chat with your data” into a workflow-focused operating layer for freelancers.

Video Demo

Video walkthrough: https://www.loom.com/share/a987e0fa63f9489fb758cbb71854868c

Show us the code

Code repository: https://github.com/VeraStorozhenko/freelance-manager

How I Used Notion MCP

This project uses the official Notion MCP server as part of the live workspace layer.

The app includes an MCP connect flow with OAuth + PKCE, so the user can connect their Notion workspace and let the copilot work with live MCP tools instead of only local app context.

That matters because it makes Notion an active operational memory layer, not just a storage backend.

When MCP is connected, the product can:

use live workspace context during copilot interactions
stay aligned with the user's current Notion state
support a more authentic AI workflow inside the product

I used MCP in a way that supports the core product thesis:

Detect -> Explain -> Recover -> Write Back

That shows up most clearly in Project Rescue Mode, where the app uses workspace context to help recover risky work before it slips further.

Tech Stack

Frontend: React + Vite
Backend: FastAPI
AI layer: Gemini
Workspace layer: Notion API + official Notion MCP server
Testing: Python unittest
Styling: custom CSS

Why I think this stands out

I did not want to build another generic “chat with Notion” experience.

I wanted to build a workflow product with a clear operational point of view.

Freelance OS Copilot is built around:

detecting project risk
explaining what is going wrong
helping recover the work
supporting direct action in the same interface

That is the difference between an AI wrapper and a usable system.

DEV Community: Вера Махлова

SteeplyOpinionated: A PR Review Tool for Tea Recipes (Returns 418, Obviously)

What I Built

Demo

Code

VeraStorozhenko / steeply-opinionated

How I Built It

Prize Category

Freelance OS Copilot: A Notion MCP-Powered Operations System for Freelancers

Freelance OS Copilot: A Notion MCP-Powered Operations System for Freelancers

What I Built

Video Demo

Show us the code

How I Used Notion MCP

Tech Stack

Why I think this stands out