4 live 1 building on the bench: Pehra Pro, Snap Receipt last shipped

What does an engineer do when the code writes itself?

Decides what's worth writing. Then reads what comes back.

I'm Muhammad Ahmed, a product-minded engineering lead and AI-native software engineer. I've shipped software since 2012: native mobile apps, backends, IoT, and platforms that real businesses run on. These days agents write most of my code. I design the systems they work inside, write the specs they work from, and review what they hand back.

Right now that means turning MVPs into secure, scalable software real customers can trust, and building AI-native workflows for product teams.

Everything on this site is either running or being built, and the Projects page says which.

Muhammad Ahmed

Ask my agent

I'm probably in a diff right now. My agent has read everything on this site and answers questions about my work, with a link to the page each answer came from.

On the bench this week

  • Building Pehra Pro

    Decoding Teltonika IO per device model, because the same IO number means different things on different models.

    last shipped

  • Live Snap Receipt

    Version 26.10.1 on both stores: the client report up front, and receipt reading rewritten for trade receipts.

    last shipped

Recently shipped

  1. PACKAGE

    libtailcat

    Tailcat for Dart and Flutter: a direct, encrypted line between two devices, with plain Dart sockets back.

  2. ARTICLE

    Sunday Brunch and Tailcat

    A Flutter library for direct, encrypted device-to-device links, built and published in one Sunday, and what that says about the cost of code.

  3. ARTICLE

    Write It Down First

    How I build with coding agents, spec first, and the remote teams that taught me the habit.

  4. ARTICLE

    Conductor vs PRCHD: Move the Agent, or Move the Desk?

    Both run Claude Code and Codex in parallel worktrees. They split the moment you leave the desk.

Running in production

  • Live PRCHD

    Story to pull request from your phone, with AI coding agents on a computer you own. Slack for everyone else.

    macOS host · iPhone · iPad · Android

  • Live Snap Receipt

    Photograph a receipt. Every line item, costed to the right job.

    iPhone · Android

  • Live QuickPeek

    Press Space in Finder: Markdown, JSON, YAML, CSV and Mermaid, rendered.

    macOS

  • Live 247 Track

    A tracking company's app for its customers: live location, trips and engine commands.

    iPhone · Android

Building:
Pehra Pro
Unreleased:
Sqwrl, FieldAudit Pro, LaunchLens, CoachTrack, ShipLock
Released:
guided_tour, libtailscale, photo_markup

Where it started

A school computer lab and GW-BASIC. Then digital logic, assembly language, and in 2006 a PIC16F887: 8 bits, 368 bytes of RAM, programmed in C. In 2010 I built a commercial GSM vehicle tracker around one.

After that came Android in Java, iOS in Objective-C, remote teams from 2014, Flutter before 1.0, and now agents.

Code got cheap. Judgment didn't.

  1. The harness is the product. A model is an engine, and nobody ships an engine.
  2. Agents are the new traffic. Backends get heavier, frontends get thinner.
  3. Agents stop waiting to be asked. People stop typing and start approving.

I used to be limited by how much code a team could write. Now I'm limited by the ideas in my head.

And AI credits. Mostly AI credits.