Mini-BQ
From WhatsApp chaos to a structured renovation management platform — built solo in one month.
Renovation ran on
WhatsApp and memory
A year-long residential renovation in Malaysia. Three parties — homeowner, main contractor, and subcontractors — each communicating in separate WhatsApp groups. Progress photos buried in chat history. Claims raised verbally or by Excel. No way to tie a photo to a specific work item, no documentation trail from site to invoice.
The Bill of Quantities existed as a PDF or Excel printout — a static document the homeowner couldn't meaningfully interact with. There was no way to know, at any point, what percentage of the contracted work had actually been completed versus what the contractor was claiming had been completed.
Disputes over defects were especially painful. Without a timestamped, location-pinned record, a defect raised verbally could be denied, forgotten, or simply not prioritised. Payments had no formal hold mechanism — pressure was entirely social.
A structured platform
homeowners can own
Mini-BQ is a web application (PWA) that gives homeowners a single platform to manage the full lifecycle of a renovation project — from importing the bill of quantities, to tracking daily work done, to processing contractor claims, to logging and resolving defects.
The system is designed for non-technical homeowners. Every piece of documentation — photos, approvals, defect logs, invoices — is tied to a specific zone on the floor plan and a specific line item in the BOQ. Nothing floats loose.
Architecture & Build
The platform is a full-stack web application built on a modern, cost-efficient stack — deliberately chosen so the entire system runs at near-zero cost during early usage, with a clear path to scale.
Before & after
The platform is live and in active use across two renovation projects, tracking a combined 324+ BOQ line items across a full year-long build cycle — from demolition through to finishing works.
| Dimension | Before | After |
|---|---|---|
| Progress tracking | Progress updates in WhatsApp — no structure, photos buried in chat | Every work done entry logged with photos, zone, and linked BOQ items. Full audit trail. |
| BOQ accessibility | BOQ as a static PDF/Excel — no way to track completion status | Live BOQ with In Progress / Completed status per item. AI OCR import from any document format. |
| Contractor claims | Contractor claims raised verbally or by email — no supporting evidence | Invoice claims generated from approved work done entries only. Evidence-first invoicing. |
| Defect management | Defects raised informally — no record, no accountability, no payment leverage | Defects pinned to floor plan with severity, photos, and assignee. Open defects documented basis to hold payment. |
| Floor plan usage | Floor plan as a paper reference only | Interactive floor plan with drawn zones. Every entry — photo, work done, defect — geolocated to a zone. |
| Financial visibility | No visibility on total claimed vs approved vs paid | Dashboard shows real-time financial summary: Total Claimed · Approved · Paid · Outstanding. |
What building solo
with AI taught me
AI is a multiplier, not a replacement. The productivity gain from Claude Code was real — a solo engineer delivering a full-stack platform in one month is evidence enough. But the quality of the output was entirely dependent on the human reviewing it. Architecture decisions, security patterns, performance trade-offs — AI generates plausible code fast. Knowing whether it's the right code requires domain experience. Human-in-the-loop is not a safety net; it is the engineering practice.
Context management is a first-class engineering problem. AI sessions have limits. Designing sprints with clean boundaries — minimal cross-dependencies, clear interfaces between modules — is not just good software design; it is the prerequisite for effective AI-assisted development. Obsidian as a knowledge graph gave the AI a reliable index to orient from, without needing to reconstruct full context from scratch each session.
Enterprise-grade is a mindset, not a team size. Security scanning, E2E testing, CI/CD pipelines, structured deployment — these practices don't require a team. They require discipline. The same standards applied at a large industrial group with a team of 20 applied here with a team of one. The constraint changed. The standard didn't.