Welcome — why this blog exists
A working log on spec-driven AI for working professionals. What you'll find here and why it's worth your reading time.
Raju H
AI Builder · Dhaka
TL;DR
This is a working log on building with AI the way working professionals actually need to — spec-first, iteratively, in production-shaped contexts. Posts are short, opinionated, and grounded in real code or real situations from the bootcamp.
What this blog is
The bootcamp covers a specific way of working with AI — write the spec, then prompt; treat AI output as a draft, not a finished artifact; design for failure modes upfront. Five days isn't enough to internalize all of it. This blog is the long tail.
Three kinds of posts you can expect:
- Methodology — the spec-driven approach, in detail. Why a spec beats a prompt. How specs differ across IDE coding, research, customer support, internal tooling. What spec-driven looks like for non-engineers.
- Tool deep-dives — Claude Code, Cursor, ChatGPT, n8n, others. Honest comparisons, including where each one breaks. Not a sales pitch.
- Case studies — Bangladesh-specific projects (bKash integration, BD startup tools), work from past bootcamp participants, my own indie builds.
What this blog isn't
Not "what is AI" content — there are thousands of those, all better than I could write.
Not vendor PR. If a tool ships a great feature, you'll hear about it. If it ships hype that doesn't survive a real production check, you'll hear about that too.
Not a marketing funnel disguised as content. Posts here exist because the underlying ideas are useful. The bootcamp link in the footer is a way to go deeper if you want to; the post is complete without it.
Why bother with Bangladesh-specific examples
Two reasons.
First, payment integrations, banking flows, government IDs, and operational tooling here have their own shape — bKash isn't Stripe, NID isn't passport, and the "send via WhatsApp" expectation runs through the entire UX. Generic AI examples don't account for any of that.
Second, the market signal matters. Bangladesh has a fast-growing tech professional layer (engineers, analysts, ops leads) that the global AI-content ecosystem mostly ignores. The bootcamp positions for that audience; this blog does too.
Reading time per post
Aiming for 5–10 minute reads. Long enough for actionable detail; short enough you can finish on a commute.
If a topic genuinely needs more, it gets split into a series.
Cadence
One post per week to start. More when the case studies pile up.
If you want to be told when new posts go live, subscribe to the RSS feed at /blog/rss.xml. For the bootcamp itself, the homepage has the wishlist signup.
— Raju H