Notes from building Luminik and earlier products: product, sales, AI-agent workflows, event pipeline systems, and the decisions that still need a founder in the loop.
Alfred turns repeated company work into plans, summaries, PR notes, and drafts I can review. This is how I keep the work short, sourced, and in my hands.
Alfred, CLAUDE.md files, isolated PRs, approval rules, and the recurring work a small company has to keep visible.
Sales starts by understanding how the buyer does the work today. Useful before hiring a sales motion.
How event ROI breaks across handoffs, and why the answer starts with one event operating record.
Mainteny, Aura, SnowOptix, and the product decisions that changed what I built next.
Alfred turns repeated company work into plans, summaries, PR notes, and drafts I can review. This is how I keep the work short, sourced, and in my hands.
Why I moved Luminik's recurring agent work into a dedicated environment, the rules that keep it under control, and what it took three upstream bug reports to learn.
I have run my own agent harness since mid-2025: role prompts, CLAUDE.md files, a specs repo, isolated branches, and PR review. What gstack solves, where the coordination gap begins, and what stays a founder's problem.
The cost structure of solo product engineering in 2026. Tools, instruction files, parallel agent work, and the judgment work that replaced what a five-person team used to do in 2014.
I delivered Luminik manually before there was a product surface. The first months inside a real GTM team taught me which parts of event work repeat, where judgment lives, and where the product boundary should sit.
B2B event ROI is fuzzy because the evidence breaks across attendee lists, enrichment, sequencer, booth, and CRM. The fix is one event record that keeps evidence attached through every handoff.
SnowOptix solved a real Snowflake cost problem, earned useful validation, and still showed the limits of the category. The conversations it created led to Luminik.
Technical founders sell well when they treat discovery as diagnosis first. Lessons from Mainteny, Aura at Bain, and Luminik on running sales calls like an engineering loop.
Aura went from zero to $3.6M ARR inside Bain in 18 months. What the firm gave us, what it cost us, and how to decide whether corporate venture building is the right setting for your problem.
A practical account of the CTO's role in an early raise: customer evidence, demos, investor fit, diligence, and keeping the product honest while the round is moving.
What it took to get field service management software for elevator maintenance teams into real customer use: narrow scope, direct sales calls, a practical stack, and enough field learning to explain the company.