Code Ships. Judgment Scales.
I'm Ramesh Rathnayake — Co-Founder and CTO of Fcode Labs , and part of the team building Forevate . Most of what ends up on this blog comes from the same place my day job does: production systems, architecture trade-offs, and the judgment calls that only show up once code has to survive contact with real users.
Fcode Labs started in 2018, when Buddhishan Manamperi, Tharindu Malawaraarachchi, and I — after a few startups that didn't pan out — decided to bet on Sri Lankan engineering talent one more time. Today we build for clients across healthcare, manufacturing, and security in seven countries. Forevate, our newest venture, is a real-time market intelligence platform built for teams who need to act before the data goes stale.
Every roadmap call starts with what the engineering can responsibly ship and support at scale.
What started in a single Colombo office now runs out of Colombo and Singapore both.
Defining what must remain true, choosing sufficient evidence, and deciding what is safe and worthwhile to ship can constrain delivery even without AI. I'm testing a related question: when AI shortens code authoring faster than validation, does verification judgment become the main delivery constraint?
Most of the evidence I've explored so far concerns review, security, reliability, and release acceptance. Architecture and engineering leadership are other ways to test the idea, not settled conclusions.
I don't think it holds everywhere, and I'd rather find where it breaks than defend it. If you've seen a case where more AI-generated code didn't shift the bottleneck to verification, tell me about it .
Start with three connected essays: whether AI makes developers more productive, what passing tests can tell us about shipping, and what vendor security evidence can actually support. These include older essays selected for relevance, not recency.
An insightful look at developer productivity with AI: exploring what the Copilot, METR, and DORA studies truly measured, and sharing how you can test these gains with your own team.
Read full dispatch
A verification contract names the claim, method, threshold and decision before you write the check, so a passing test proves what the release decision needs.
Read full dispatch
A vendor security questionnaire proves a dated self-report, not a working control. See what it really answers and which checks you still owe afterwards.
Read full dispatch