Product manager, founder, builder.
I'm an independent product manager, founder, and software builder based in Kathmandu. I help teams make product decisions and adopt agentic software development, drawing on over a decade of building software. I write about product thinking, engineering with AI, and hands-on experiments.

The journey
Engineer, founder, now independent.
Over a decade building software. Selected chapters from engineering, founding, and consulting.
Software Engineer
Leapfrog Technologyยท2014 โ 2017
Three years across Rails, React, and React Native. Trainee โ Technical Project Manager.
Founder
Truemarkยท2017 โ 2025
Founded and ran a product engineering studio for eight years. Wound it down to focus on a single product team.
Contract Software Engineer
Luthor (YC F24)ยทPrevious engagement
Worked with Luthor as an independent contract software engineer.
Product manager & consultant
NowIndependentยทNow
Freelance product management, codebase readiness audits, software factory setup, and agentic coding adoption.
Now
What I'm focused on this season.
Updated regularly ยท inspired by nownownow
Parallel agent coding
OpenSpec for spec-before-code, Beads as the local task graph, git worktrees for isolation. The toolkit I wish I had.
Updated May 2026
Hermes for release notes
Wiring a Hermes agent to draft release notes from commits and PRs. Cron-scheduled, model-agnostic, ~$5 of infra.
Updated May 2026
Field notes ยท AI & agentic engineering
On building software with AI agents.
New workshop ยท free, self-paced
Stop repeating yourself to your agent.
A hands-on workshop that turns the four-layer agent-ready stack into muscle memory: one prompt, one hook, one subagent, one skill, against a planted bug in a starter repo.
Hook
Block the diff before the agent finishes.
Subagent
Audit the rest of the repo for siblings.
Skill
Name the workflow so you can call it.
Listen
Latest essay, in your ears.
Audio companions to selected essays. Generated, then loudness-normalized and published as a GitHub release.
The Subagent I Didn't Need
A confession about a subagent I wrote, ran for a month, and quietly deleted โ and the broader taxonomy of agent-side artifacts that helped me see why. Five things live in a mature agent-ready repo, and they are not interchangeable. A field guide to when not to delegate.
Owning my stack ยท 2026
I swapped a music subscription for a stack I own
AI changed the maths for everyday software. When a coding assistant can help write the glue, running your own small stack starts to beat a made-for-everyone SaaS. My music setup is the test case: a lossless library on Navidrome, AI-powered "more like this" mixes, listening stats, and offline listening on my iPhone and an old iPod.
Scripts, not subscriptions. With an AI coding assistant, a sync script or a cleanup job can be an afternoon, not a project.
Own the stack. My files, my server, my history. Nothing disappears when a licence or a price changes.
Keep the fun. Sounds-like mixes, listening stats and a yearly recap, without the monthly bill.
Reading paths ยท In-depth series
Pick a track. Each one ships a complete idea end to end.
Tech notes
From the engineer's notebook.
Patterns, debug stories, and small wins โ written for my future self.
Browse all
Assistant-Only Work: Calendar and Inbox on Potts

One Skill Registry for Every Harness

Steering Hermes Agents From One Place

Assistant-Only Work: Calendar and Inbox on Potts
Inbox triage, a morning digest and calendar management run only on my assistant instance. Nothing else needs those skills or that Google grant, so nothing else gets them.

One Skill Registry for Every Harness
Claude Code, Codex, Hermes and Pi get their agents and skills from one registry. Upstream skill packs are pinned to exact commits and never patched; changes are forks with provenance.

Steering Hermes Agents From One Place
One shared room, @mentions and the Hermes web and mobile apps let me redirect four agent instances mid-task. The Google Photos to NAS migration shows how that works in practice.

Four Hermes Instances, One Model
Engineering, product, growth and a personal assistant run as four separate Hermes instances on the same model. Each has its own VM, profiles, skills and boards, and a narrow way to reach the others.

Agents Aren't Cron Jobs With a Personality
Cron decides when work starts. An agent decides what happens next. Why I run four Hermes instances, and why the harness around a model matters more than the schedule that wakes it.

Building My AI Shed: A 10G Server Backbone and 160 GB Across Two Minisforum Hosts
Why I moved my homelab to a 10G server backbone: an always-on AI shed on my tailnet, where agents, builders and local models share fast storage. Includes measured throughput.

Letting Coding Agents Build It: An Orchestrator, Parallel Lanes and Verification Over Trust
Much of the Campsite work in this series was implemented by coding agents in parallel worktrees, coordinated by another agent. What made it work was re-checking every result before it merged.

From MySQL to PostgreSQL with Patroni HA
Campsite now runs on PostgreSQL with Patroni HA across two hosts. Getting there meant surfacing MySQL behaviour the code never mentioned, running one test suite on both databases, and replacing pgloader with a small Ruby copier.

Deploying Campsite with Kamal: Split Configs, Exact-Commit Images and a Secrets Preflight
One Kamal config per runtime, a background worker that can't be started by accident, images tagged with the exact commit they came from, and a secrets check that never prints a value.
Working notes
What people say after shipping together.
A few field notes from clients and peers who worked with me across product, engineering, and company-building.
Prakash and the Truemark team stand above every engineering team I've worked with over the last 25 years. If you can partner with Prakash, you will not regret it.
I have seen him progress from Trainee to Technical Project Manager. He is an excellent software developer, very enthusiastic, a quick learner, and someone with excellent problem solving skills.
Prakash is really easy to work with, and has a good understanding of react-native and javascript. I hope to work with him again one day.
Prakash and the Truemark team stand above every engineering team I've worked with over the last 25 years. If you can partner with Prakash, you will not regret it.
I have seen him progress from Trainee to Technical Project Manager. He is an excellent software developer, very enthusiastic, a quick learner, and someone with excellent problem solving skills.
Prakash is really easy to work with, and has a good understanding of react-native and javascript. I hope to work with him again one day.






