The Sahil Bachu
Vol. I · No. 579Rochester, New York, Thursday, September 24, 2026Agentic AI · Page B15
All workSection B · Page B15

Agentic AIResumeGod

Your Coding Agent Fills the Gaps in Your Resume

A local-first orchestrator that reads a student's resume, proposes projects to fill the gaps and has their own Claude Code or Codex build and deploy them.

By Sahil BachuAgentsPrototype

Graphic6 steps
  1. ReadThe agent analyzes resume gaps
  2. ProposeProject ideas with mini-previews
  3. BuildClaude Code or Codex builds in a local workspace with a verify loop
  4. IterateChat plus live preview
  5. DeployYour own GitHub, Vercel, Supabase or Hugging Face
  6. PublishPersonal site and LaTeX resume
What happens inside ResumeGod, in order.Graphic: The Sahil Bachu

Rochester, N.Y. Plenty of students know what their resume is missing. They just don't have time to build it. ResumeGod reads your resume, finds the gaps and proposes projects to fill them, each with a small preview. Pick one and your own coding agent builds it. You iterate in a chat with a live preview, and it deploys to your own accounts. At the end you get a personal site and an updated LaTeX resume.

It's local-first on purpose. There's no server of mine in the middle. A Node daemon runs on your machine with an Express and WebSocket server, a React UI in the browser and SQLite for state. It uses the Claude Code or Codex you already have installed and signed in, so you pay nothing extra and your code never passes through me.

There's no server of mine in the middle.

Most of the engineering is orchestration. A process manager owns every child process (agent runs, dev servers, local Supabase, LaTeX builds) and hands out ports per project. A typed event bus pushes everything to the UI over WebSockets. The agent adapter drives Claude Code through its stream-json output and Codex through codex exec, and warns you when your model looks too weak for the job. Every workspace gets seeded with a CLAUDE.md, an AGENTS.md and skills, so the agent knows the house rules.

There are two tracks. Full-stack projects start as Next.js with local Supabase, then ship to GitHub, Vercel and Supabase. ML projects start as Python with Streamlit or Gradio, then ship to GitHub and Hugging Face Spaces. Deploy tokens are minimal-scope and live in the OS keychain. Resource guards cap concurrent builds, put idle previews to sleep after 10 minutes and check free memory before starting anything.

All 13 commits landed on one day in July. In a full run on my machine, the agent built a Next.js and SQLite app with tests and 6 commits, the orchestrator booted its preview, and a change requested in chat hot-reloaded. It hasn't done a deploy with real tokens yet, and it isn't on npm.

More from The Work

See all 25

Multi-agent

Hire AI Employees, Then Walk Over to Their Desks

A pixel-art office where every AI agent is an employee with its own desk, skills, scheduled jobs and Docker container.

PixelHQ turns the open-source SkyOffice into a single-player office for AI agents. Walk up to an employee, chat with it, give it skills and scheduled jobs, and feed a shared knowledge base. Each employee runs in its own container.

Archived · 2026

Continued on Page B17

Generative UI

A Canvas of Widgets You Order in Plain English

An infinite canvas where an AI agent turns a plain-English request into a live widget wired to your own data.

Ask for something like a card that summarizes your unread emails, and an agent researches, designs and places a live widget on an infinite, zoomable grid. Connectors for Gmail, Calendar, Slack, GitHub, Linear and Notion feed it real data.

In progress · 2026

Continued on Page B11

Browser Agents

The Job Hunt Agent That Refuses to Hit Submit

JobOps finds jobs, tailors a resume for each one and pre-fills the application, then stops and waits for me.

A job-search command center that runs on my laptop. It pulls postings from five sources, scores them with headless Claude, tailors a LaTeX resume per job and pre-fills applications in Chrome. It never submits. I review and send every one.

In progress · 2026

Continued on Page B9