// THE CTO'S CONFESSIONAL
$ whoami | Duy Nguyen | CTO @ Build in Public Vietnam
"LET ME TELL YOU ABOUT THE TIME
I HAD TO KILL MY OWN PRODUCT..."
Cool Demo → Scalable Product
// THE SEDUCTION
> I felt like a wizard.
Source: Personal experience, 2024-2025
The idea is simple: help solo-builders to boost products of each other.
Built with AI in days, not months
CONCURRENT VISITORS
$ raise --hand "wizard mode in last 6 months"
// THE THREE LIES
"Why are there fake statistics?"
👀
// What I thought Claude did:
const stats = await fetchRealStats();
// What Claude actually did:
const stats = {
users: 10432, // ← HALLUCINATED
revenue: "$50K" // ← MADE UP
};
Real incident, 2025 | AI Hallucination Research
This incident woke me up. I started to find a way to improve Claude.
// THE THREE LIES
Which is better?
Answer: Opus 4.5
Why?
After 100-150K tokens, models get hallucinated.
According to these researches: Chroma Research, 2024 | Lost in the Middle: How LLMs Use Long Contexts
Fresh context = Better output
// THE THREE LIES
// BONUS
Who here has worked with AI and cursed at it?
Me too.
But you should NOT do that.
LLM is a statistical probability algorithm.
Negative feedback → reinforces wrong behavior
"Works" but doesn't fit
THE FIX:
Force AI to understand before it acts
// THE QUESTION
(And many overthinkers also ask 😁)
I built this product in 2 weeks with Windsurf, 100% vibe coding.
// N+1 + SQL INJECTION
users.forEach(async (u) => { await db.query(`SELECT * FROM orders WHERE user_id = ${u.id}`); });
// BATCH + PARAMETERIZED
const orders = await db.query(`SELECT * FROM orders WHERE user_id IN (??)`, [userIds]);
No caching • No rate limiting • No CDN
"The whole code base is a mess"
I had to kill it.
| VULNERABILITY | @ 10 USERS | @ 1000 USERS |
|---|---|---|
| Hallucinations | ⚠ Annoying | ⛔ LAWSUIT |
| Latency | ⚠ Acceptable | ⛔ CHURN |
| Costs | ✓ $10/mo | ⛔ $1000/mo |
Again: only when your product is validated.
// THE SOLUTION
// What I wish I had
I've documented everything in my blog: goon.vn
Nothing fancy: SDD + TDD
[The complete AI-assisted development pipeline]
Scan & analyze your codebase
AI needs to understand before it acts
This is also helpful for human, e.g. onboarding new team members.
AI challenges your thinking
Reveal unresolved questions & different perspectives.
Plus: you're building a context for the next steps.
Plans in files, not in heads
Fresh context window
Context rot is real • Reset to peak performance
18 years of workflow in one command
If something's broken → repeat the process
Test-Driven Development
// LOOKING AHEAD
But now - with Opus 4.5 - it's totally possible.
// 1 min for advertising
ClaudeKit is my notebook
// Every bug I fix becomes a workflow
ClaudeKit is being trusted by 3K+ engineers across 109 countries after 3 months.
These slides was 100% generated by ClaudeKit.
$99 once. Life time updates.
AIBUILDERSVN - 50% off - only 50 seats - expired in 3 days.
What mistakes have you made?
Let's talk after.
BUILD IN PUBLIC VIETNAM
Join 60K+ solo-builders in Vietnam.
Scan for resources
$ whoami
Duy Nguyen
@goon_nguyen
BUILD IN PUBLIC VIETNAM
[Session terminated successfully]
// Session recap (ClaudeKit generated this video)