My 8 months.
aitax — a tax automation platform for Singapore. I committed, iterated, raised money, learned to sell. The thesis I started with was wrong.
17 ideas. 8 criteria.
You think they're all good. But you don't know which one to work on. I gave Claude 8 criteria based on my skills, my market, my situation. One session. Ranked list. Not guessing — data.
| Product | PMF | Fit | AI | Mkt | Fut | VC | Rev | Scl | Total |
|---|---|---|---|---|---|---|---|---|---|
| P11: AI Dev Team | 1 | 3 | 3 | 2 | 3 | 3 | 2 | 3 | 20/24 |
| P10: AI Associate | 1 | 3 | 3 | 2 | 3 | 2 | 1 | 3 | 18/24 |
| P0: Task Automation | 2 | 3 | 3 | 2 | 2 | 1 | 3 | 1 | 17/24 |
| P12: BidCommand | 0 | 2 | 3 | 2 | 3 | 3 | 1 | 3 | 17/24 |
| P13: ValuCommand | 0 | 2 | 3 | 2 | 3 | 2 | 1 | 3 | 16/24 |
| ... 12 more products scored | |||||||||
Where's the data?
Singapore has 5,000+ public datasets. 100+ API endpoints. Free. Commercially usable. Claude mapped all of them to my 5 products in one session.
# 5 parallel research agents + Council debate
APIs DISCOVERED: 100+ endpoints across 15 agencies
DATASETS: 5,000+ from 65+ agencies
LICENCE: Free. Commercial. Perpetual.
KEY FINDING:
One data.gov.sg API key powers
4 out of 5 products.
GeBIZ • URA • OneMap • MOM • NEA • BCA
ACRA • LTA • SingStat • PUB • EMA
Each idea gets a home.
Claude builds the whole thing. Product spec. Outreach plan. Contact list. Personalized emails. A working demo. All from one conversation.
12-bidcommand/
├── README.md ← what is this product?
├── spec-outreach-list.md ← who do I reach?
├── outreach-list.csv ← 100 real contacts
├── outreach-emails.csv ← personalized drafts
├── gebiz-data.json ← real API data
└── intel-tool/ ← working demo
├── index.html
└── data.json
What is BidCommand?
Not a vague idea anymore. It's a product spec. Who buys it. What it costs. What the demo looks like. Claude wrote this from our conversation about the market.
> Type a tender number.
> Get a draft Technical Proposal.
> Edit, submit, win.
Market: SG gov procurement ~S$60B/year
Buyer: Bid Managers at mid-size companies
Price: $5,000/month
Pain: Each tender takes 1-4 weeks to write
DEMO CLOSER:
"Type your competitor's name.
See every tender they won since 2020."
Who do I reach?
Claude writes the spec for finding prospects. Not vague — it tells me exactly where to look, how to filter, what to export. I just run it.
Phase 1: Company Discovery
Query GeBIZ awarded tenders API
Filter mid-size companies (200-500 staff)
Rank by bid frequency
Phase 2: Contact Enrichment
Search Apollo for Bid/Tender Managers
Pull: name, title, email, phone
Target: 2-3 contacts per company
Phase 3: Clean & Export
Dedup, verify, export → outreach-list.csv
100 contactable prospects ✓
Give value first.
Before I even meet a prospect, I send them a link. "Type your competitor's name." They get real intel — for free. Now they want to talk.
Search: CrimsonLogic
18,000+ awarded tenders indexed
Tenders won: 12
Total value: $14.2M
Agencies: MOE, MHA, IMDA, GovTech
Competitors: NCS, Accenture, DXC
→ Free. No login. Send before the meeting.
Now multiply by 5.
I'm not doing this for one idea. I'm doing it for five. In parallel. Same day. Like a developer with 5 terminals open — but instead of debugging code, I'm debugging product-market fit.
The battle card.
When I get a meeting, Claude generates a battle card. Time-boxed agenda. Objection handling. Exact phrases. Hormozi's frameworks loaded as a skill. I walk in prepared.
# Goal: Lock scope + get payment
# Time limit: 20 min. Not another hour.
Phase 1: Walk through scope (8 min)
Phase 2: Confirm requirements (2 min)
Phase 3: Payment discussion (3 min)
Phase 4: Get homework (3 min)
Objection: "Can you do it cheaper?"
→ "This is exactly what you described.
$X now, $X on delivery."
Framework: Hormozi — Value Equation
$5,000.
5-6 paying clients. 30 conversations. Each one built in a single session. Tax reports, sales pipelines, document generators. The system works.
Try my tools.
Open source Claude Code skills I built along the way. Install any of them in 10 seconds.
| Skill | What it does |
|---|---|
| naval-explain | Visual learning with multi-model explanation |
| balaji-slides | Evidence-driven slide design (like this deck) |
| x-poster | X/Twitter post optimization with quality scoring |
| spec-driven | Brain dump → interview → spec |
| hormozi | Alex Hormozi's $100M business growth system |
x.com/phuakuanyu
ns.com/kuan/apply
