Think about year three
The question is not what is quickest to build now, but what a different developer can safely change in three years. That favours well-known tools.
Proven tools, chosen on purpose
Anything unusual is a cost, and it is paid by whoever maintains your system after us. We use well-established tools with large communities and people you can actually hire.
Selection principles
How a technology gets chosen
What we reach for most often is below. Which part applies to your project is decided by these six questions, roughly in this order.
Keep it as simple as it can be
A normal database until the data genuinely needs something else. Complexity has to earn its place, because someone pays for it later.
Stay able to change your mind
AI models especially. We build so you can switch provider without a rewrite, because today’s best choice will not be next year’s.
Fit what you already run
If your team already works in a particular stack, we are unlikely to introduce a second one for the sake of it.
Security comes first, not last
We check how well a dependency is maintained before adopting it, and scan for vulnerabilities on every change.
No kickbacks
We hold no reseller agreements or partner commissions with anything on this list. Recommendations reflect your situation, nothing else.
The stack
What we build and run on
Languages
Widely used, so you can hire for them later.
- TypeScript
- JavaScript
- Python
- SQL
Web
Fast, simple pages by default. Heavy front ends have to justify themselves.
- React
- Next.js
- Astro
- Angular
- Svelte
- Node.js
- Express
- FastAPI
- htmx
- Tailwind CSS
Mobile
One codebase for iPhone and Android, with offline support where it is needed.
- React Native
- Expo
- Ionic
AI & LLM
Built so the model can be swapped without a rewrite.
- Claude (Anthropic)
- OpenAI
- RAG
- LangChain
- Pydantic AI
- Pinecone
- Voice bots
- Dialogflow
APIs & automation
Clear, documented connections between the systems you already use.
- Python
- FastAPI
- REST APIs
- GraphQL
- Swagger / OpenAPI
Data & reporting
Picked to fit the data, then kept as simple as possible.
- Firebase
- MongoDB
- BigQuery
- DynamoDB
- Looker
- Power BI
Where it runs
Your accounts, not ours. We set it up and hand over the keys.
- Google Cloud
- Cloud Run
- AWS
- AWS Lambda
- Docker
- Terraform
- GitHub Actions
A note on AI tooling
Model choice should stay reversible
AI moves faster than anything else here. The best model today may not be the best in six months, and prices change just as quickly.
So we build so the model can be swapped without a rewrite, keep your accuracy tests independent of any one provider, and track cost per request. Where your data rules require it, the same setup runs an open-source model on your own infrastructure.
Our LLM development serviceWant a second opinion on your setup?
Tell us what you are running and what is causing you trouble. We will give you a straight view on whether it needs changing — and often the answer is that it does not.
We reply to every enquiry within one working day.