LATEST INSIGHTS
Deep dives into technology, development, AI trends, and the thinking behind what we build at Code(Hard);.
Treating Thailand's PDPA as a checklist you complete before launch produces brittle systems. We treat it as an architectural constraint — consent, data lifecycle, and access control are part of the data model from day one.
Thai is written without spaces between words. Most global search stacks treat a whole sentence as one token — so a search for a keyword inside a document silently returns nothing. Here is how we segment, index, and rank Thai text correctly.
Data residency rules, procurement constraints, and day-two operations all pull in different directions. Instead of a religious debate, here is the framework we use to decide where a public-sector workload should actually run.
Editors work in a CMS backed by Firestore; visitors get plain static HTML with no client-side data fetch. Here is how we get the SEO and speed of static generation without giving up a live editorial workflow.
A scanned circular is invisible to search until OCR turns it into text. Thai OCR has its own failure modes around layout, scan quality, and tone marks — here is what it takes to turn a pile of scans into a searchable archive.
Most systems log something, but a trail nobody can reconstruct is just storage cost. Here is what separates an audit trail that satisfies a regulator from one that becomes a liability when the questions get specific.
ThaID gives your application a legally recognized digital identity check backed by DOPA. The protocol is straightforward on paper but has sharp edges in practice — here is the flow, the common pitfalls, and how we handle sessions afterward.