Services: build and ship
We still build the systems and websites, now with AI in the dev loop.
We also build the plumbing, the marketing sites and the search content that brings in traffic, in-house, on the same .NET, Node, Postgres, MS SQL, AWS and Azure stack we've used for eighteen years. What changed in 2026 is that AI tooling sits in the dev loop, and most new products include at least one agentic feature.
Product engineering
The stack we've shipped on since 2008, with AI in the dev loop and in the product where it's useful.
Stack
.NET 8/9, Node 20+, Postgres 16 and MS SQL Server 2019/2022, on AWS, Azure, AliCloud and Huawei Cloud. We work where you already operate.
What changed in 2026
A senior engineer works with agentic CLIs such as Claude Code, Codex CLI and Gemini CLI in the dev loop. Scaffolding and review are faster, and there are fewer regressions. The product still ships as tested code through a tested pipeline, and a named person approves each merge.
Agentic features
Most new builds now include at least one narrow agent: a copilot, summariser, structured extraction or classification. We design the safety envelope and the eval harness before the agent reaches a user.
Motion.astro+78
-
Check written first
$ python3 check_motion.py reveal: below-fold card starts at opacity 1, expected 0 counter: 400+ never climbed
-
Code, then the same check
$ npm run build [build] Complete! $ python3 check_motion.py ALL CHECKS PASSED
-
Human review
“current one is good. keep it. commit.”
-
Committed to the branch
$ git commit c8b535c feat: add scroll motion and stack CTO comparison on phones 4 files changed, 128 insertions(+), 1 deletion(-)
Corporate and product websites
Marketing sites and product websites, on infrastructure we operate.
What we deliver
Design system and tokens, site implementation, headless CMS or MDX-driven content, and deploy infrastructure. We use Astro for marketing sites and Next.js for app-like products. Typography and layout follow a living style guide.
Infrastructure we operate
Snapshotted releases, atomic rsync deploys and rollback by release ID. If your CDN edge fails, we have a shell and a runbook, and we can reach the platform at two in the morning.
This site as an example
Its typography, taxonomy, agent team pages, Field Notes, and deploy and rollback process show what the engagement looks like on your domain.
SEO and GEO content
Content at volume that keeps your voice, from a research desk.
Article Studio does the work
Nine stages, ten specialists and a tier-graded source registry. Every claim is cited and every URL is fetched live, in English and Simplified Chinese where needed. See how it runs on the Article Crew page.
Singapore and APAC primary sources
We read SFA, IMDA, ESG, A*STAR, MOH and provincial Chinese primary sources in their own language and trace claims back to them. Local knowledge has to come from the source, and an LLM's summary of Singapore doesn't count.
Search and generative-engine optimisation
Each piece is structured so that both can retrieve and quote it, with schema where it helps, semantic structure that survives a chunker, and enough citations for an answer engine to rely on without hallucinating.
A senior editor on every piece
The editor reads every piece end to end before publishing. We refuse listicles without direct project experience and "What is [buzzword]?" explainers without an opinion. When voice and volume collide, we choose voice.
Engagement shape
A content retainer, with monthly throughput against an editorial calendar, or a one-off pillar build, with a set of cornerstone pieces around a launch. We'll recommend one during the diagnostic.
Draft under checkIllustrative
- fact-check.agent Maps every claim to a citation. Below 95% coverage, the draft goes back to the writer.
- fact-check.agent Checks every quote verbatim against its source.
- reference-verifier.agent Resolves DOIs at Crossref and fetches every URL live.
- Editor (human) Reads every piece end to end before publishing.
Common questions
What people ask before they brief us.
What stack do you build on?
.NET 8/9, Node 20+, Postgres 16 and MS SQL Server 2019/2022, on AWS, Azure, AliCloud and Huawei Cloud. We work where you already operate. We use Astro for marketing sites and Next.js for app-like products.
Do you also write SEO and GEO content?
Yes, through the Article Studio crew: nine stages, ten specialists and a tier-graded source registry. Every claim is cited and every URL is fetched live, in English and Simplified Chinese where needed. You can take it as a content retainer against an editorial calendar, or as a one-off pillar build around a launch.
Do you operate the infrastructure after launch?
Yes. We use snapshotted releases, atomic rsync deploys and rollback by release ID. We stay off third-party platforms we can't reach at two in the morning, so if your CDN edge fails, we have a shell and a runbook.
Will you add agentic features, or is that a separate engagement?
It's the same engagement. Most new builds now include at least one narrow agent, such as a copilot, summariser, structured extraction or classification. We design the safety envelope and the eval harness before the agent reaches a user.
Brief us
Tell us what
you need built.
A paid one-week diagnostic decides scope, stack and engagement shape. You keep a written proposal whether or not we go further.
Brief us