Nitin Khanna/Hire

Now

What I'm building with AI agents. I run most of my own engineering through eight of them. Each has one job and a written list of things it isn't allowed to do.

Those limits are what make me trust the output enough to ship it. I wrote more about this in What my agents can't do is the point.

  1. Bootstrapper
  2. PM
  3. Architect
  4. Coder
  5. Reviewer
  6. QA
  7. Shipper
  8. APK Builder

Northline Ops and Chitta

Northline Ops was a personal experiment. I wanted to see if an agent could help me show up on Reddit more often without turning into spam. It reads a thread and the community's rules, suggests a reply, and checks it. I edit it, and I post every word myself from my own account. In about a month the account went past 4,000 karma, and Reddit's stats put the reach in the millions.

Chitta is the memory I'm building next to it. It remembers the people you talk to, so you can pick up a conversation where you left it. Together they're meant to help indie developers build a real audience before they launch, so there are people who already know them on launch day.

  1. ReadThe thread and the community's rules
  2. RouteAn adapter per community, with voice rules and a risk tier
  3. ResearchOnly if facts are missing, and I approve the searches
  4. DraftTwo drafts in my own voice, which I usually rewrite
  5. ScoreRules, reputation risk, fit and voice, from 1 to 10
  6. I postBy hand. It never posts, votes or messages anyone
  7. RememberChitta records what was actually posted and who replied

See how it works, case by case

What I've shipped recently

  • Go, open source
    DossierX

    A Go tool that turns YAML claims into a checked, linked website. Agents can write the claims, but only a person can lock them. 30 releases in 11 weeks.

  • Python, open source
    veda-mcp

    An MCP server that lets agents read public Reddit threads and web pages safely. It has bearer auth, respects robots.txt and blocks SSRF.

  • TypeScript, agent
    Tailor

    Built for neatHack in October 2026. You describe the AI agent you're building, and Tailor proposes, checks and explains a set of sample traces for it in a Neatlogs-style dashboard. Watch the demo.

  • macOS, Swift
    Curtainly

    A Mac app that notices when you're sharing your screen and hides your private apps, notifications, tabs, desktop and Dock.

  • Cloudflare Worker
    Commute advisor

    Every weekday morning it checks bike traffic and the rain forecast, then sends me one time to leave for work.

  • Android
    Earlier open source

    Instagram Profile Downloader (104 stars) and contributions to NewPipe.

Recent writing

Building something in mobile, money or AI? Email me. I reply within a day.

Email hey@nitinkhanna.io Recommendations GitHub Writing

People I've worked with at Vance, StockGro and elsewhere have written recommendations on LinkedIn.

Full time as a Staff Engineer, Head of Mobile or in applied AI. Part time as a fractional lead or advisor. Based in Bengaluru, working with teams in the US, Europe and Asia.