daniel.tsxEmail

Nguyễn Cao Thắng — online as Daniel TSX

Nine years of building interfaces. Now shipping whole products.

I'm a senior frontend engineer — React since 2017, Next.js since 2020 — who now designs, builds, and launches AI-native SaaS products end to end at Eastbase Studio, my independent product studio. The interface craft stayed; the rest of the product came with it.

Selected workEastbase Studio

github.com/daniel-tsxdanieln.tsx@gmail.com

02Currently

The current chapter

For most of my career I was the frontend specialist on someone else's product. Over the last two years I've deliberately widened that: building real SaaS products under Eastbase Studio — not portfolio demos — and owning every layer it takes to put them in front of paying users.

AI agents are a working part of that loop, not a talking point. The products are at different stages — some with paid plans live, some free, some still early — and I'd rather show that honestly than dress it up. Shipping is the practice; the studio is where the reps happen.

End to end, concretely

frontend·backend·auth·database·billing·email·storage·analytics·launch content·docs·polish

03Selected work

A few things, told properly

Not the whole catalog — that lives at the studio. These four show the range: consumer AI, trust-sensitive workflows, the infrastructure concerns that come from actually running AI products, and the foundations distilled into kits.

The studio

Eastbase Studio

eastbase.studio

A small, independent product studio from Asia — and the container for everything below. SaaS products, developer tools, AI-assisted workflows, and premium Next.js kits, built and run by one engineer. It's where I practice the full loop, repeatedly: idea, build, launch, iterate.

W.01Live

SmartTrips

Map-first AI trip planner

Turns vague trip ideas into an editable route on a real map — with itinerary saving and sharing, public templates, and a companion mode for the trip itself. It explores how AI planning output becomes a structured product surface instead of a wall of chat text.

Notes Streaming AI itineraries into map state, a data model that survives human edits, and a “Reality Check” pass that validates plans against the real world. My deepest consumer-AI surface so far.

Next.js · AI SDK · Gemini · Mapbox · Neon · Drizzle · BetterAuth

Visit smarttrips
SmartTrips map-first itinerary planner interface

SmartTrips — production interface

W.02Live

DueKind

Approval-first receivables assistant

Tracks unpaid invoices for freelancers and small businesses, drafts relationship-safe follow-ups, and routes every reminder through human approval before it sends from the user’s own Gmail or Outlook. It explores AI in a trust-sensitive workflow, where a bad message costs a client.

Notes OAuth into Gmail and Microsoft Graph, an approval pipeline where AI drafts but never sends, and audit history for every touch. Designing for “AI proposes, human decides” end to end.

Next.js · Gemini · Gmail API · Microsoft Graph · Drizzle · BetterAuth

Visit duekind
DueKind receivables dashboard

DueKind — production interface

W.03Live

BurnCap

AI cost monitoring for small teams

Attributes AI spend by feature, customer, and model, forecasts month-end costs, and enforces budget checks — deliberately outside the critical request path. Built because running AI products made the cost question personal.

Notes Ingestion and attribution modeling, forecasting from partial months, and dashboard design where the numbers must be trustworthy before they can be pretty.

Next.js · PostgreSQL · Drizzle · Serverless ingestion

Visit burncap
W.04In progress

Eastbase Kits

Premium Next.js starter kits

A kit studio that packages the foundations I rebuild for every SaaS — auth, database, billing, email, launch pages — into premium Next.js starting points, written so AI agents can work in them as effectively as humans. It explores productizing the stack itself.

Notes Distilling repeated product foundations into something teachable: conventions, docs, and agent-ready structure instead of one-off scaffolding.

Next.js · TypeScript · Drizzle · BetterAuth · Lemon Squeezy

Also on the bench

AegisRail (release guardrails)·LodeStar (AI-agent visibility)·OkayFlow (client approvals)·Cookd (work journal)·PodCut (podcast repurposing)·StackPulse (open-source release feed)

The full, current list lives at eastbase.studio
04Capabilities

What I can own

C.1

Interface architecture

Nine years of React, six of Next.js. App Router and server components, design systems, complex dashboards and editors — interfaces that stay fast and maintainable as the product grows.

React · Next.js · TypeScript · Tailwind CSS · Framer Motion

C.2

SaaS foundations

The unglamorous layer that makes a product real: authentication, relational data, billing, transactional email, file storage — wired with a serverless-first architecture that one person can operate.

BetterAuth · Neon · Drizzle · Lemon Squeezy · Resend · Cloudflare R2

C.3

AI in production

Streaming AI features that hold up outside the demo: prompt and workflow design, graceful failure, latency budgets, and cost awareness — I care enough about that last one to have built a product for it.

Vercel AI SDK · Gemini · Claude · GPT models

C.4

Launch readiness

Getting a product over the line, not just to the repo: landing pages, docs, SEO and social cards, analytics, error tracking, and the final polish passes that separate shipped from published.

PostHog · Sentry · SEO / OG · Docs

05Method

Working with agents

AI agents write a lot of code in my workflow. The value isn't that they write it — it's knowing what to build, how to direct the work, and what good output looks like. That judgment is what nine years of shipping buys.

Daily Claude Code · Codex · Cursor · Claude · GPT models

  1. M.1

    Direct, don’t delegate blindly

    Agents get real specs: constraints, existing patterns, definitions of done. Every diff gets reviewed like a colleague wrote it — because effectively, one did.

  2. M.2

    Agents across the whole loop

    Not just code. Tests, migrations, docs, launch copy, QA sweeps, review passes. The workflow is built so several fronts move at once without quality drifting.

  3. M.3

    Judgment stays human

    What to build, what “done” means, where the security and quality bars sit, when output is wrong even though it looks right. That is the actual job.

06The record

How I got here

  1. 2017

    Frontend, professionally

    First engineering roles building for the web, with React from early on. Product interfaces became the specialty fast.

  2. 2020

    Next.js becomes home base

    Went deep on Next.js — rendering strategy, application architecture, performance. It has anchored my work ever since.

  3. web3

    Three years of dapp frontends

    A chapter building web3 interfaces along the way — ethers.js and viem, wallet flows, making on-chain state feel like a normal product.

  4. 2024

    The product turn

    Expanded from frontend specialist to end-to-end builder: auth, data, billing, email, launch. The first real SaaS products shipped.

  5. Now

    Eastbase Studio

    Building AI-native SaaS products under my own studio — some with paid plans live, some free, some early — and open to selected collaborations.

07Availability

Where I can plug in

I take a small number of engagements alongside studio work — deliberately narrow: senior frontend and product builds where the craft actually matters.

  • Senior frontend / product engineering

    Embedded in a remote team, owning React & Next.js surfaces end to end.

  • AI-native product builds

    From idea to launched product — the same loop I run at Eastbase, applied to yours.

  • SaaS dashboards & product surfaces

    The complex screens: dashboards, editors, onboarding, billing flows.

  • Product polish & UI systems

    A focused pass that takes a working product from functional to credible.

danieln.tsx@gmail.comor start with the work at eastbase.studio