My Tech Stack 2026

Why I Write About My Tech Stack

Every spring I take stock. Which tools do I use daily? What has changed? What did I drop, what is new? This article is my inventory for 2026, and I hope it is useful for other developers questioning their own stack.

A disclaimer up front: there is no perfect tech stack. There are only stacks that fit specific requirements. My stack is optimized for a solo developer and small team building SaaS products, managing cloud infrastructure, and working heavily with AI. Your context is different, and your decisions should be too.

Programming Languages

TypeScript / Node.js — The Swiss Army Knife

TypeScript is my primary language for almost everything. Backend services, CLI tools, infrastructure scripts, frontend logic. The decision for TypeScript was not always clear — I come from the C# world and struggled with it for a long time. But the productivity I achieve with TypeScript and the npm ecosystem is simply unbeatable.

What convinced me: a single type system for frontend and backend. I define my API types once and use them everywhere. No code generator, no mapping, no discrepancies. Add to that the massive ecosystem of libraries and the fact that practically every developer can read JavaScript or TypeScript.

For backend services, I use Node.js with Fastify or Hono as the framework. Express was my go-to for a long time, but Fastify offers better performance and a cleaner plugin system. Hono has established itself as my favorite for serverless functions — it is lightweight, fast, and runs everywhere.

C# — The Enterprise Companion

C# is still part of my stack, but its role has changed. I use it when working with existing .NET systems or when a client has a Microsoft-centric infrastructure. The language itself is excellent — it has evolved fantastically in recent years. But for new projects, I almost always reach for TypeScript.

Python — The AI Bridge

I use Python primarily for AI/ML tasks. Not because Python is the best language, but because the AI ecosystem lives 90% in Python. When I work with LangChain, Transformers, or other ML libraries, Python is the logical choice. For everything else, I avoid it — the type system is too weak for my taste and dependency management is still a nightmare.

Cloud Platforms

Hetzner Cloud — For My Own Products

All my own products and projects — Atoo Studio, SiteHorse.ai, the IT-Trail SaaS Stack, Network Tools — run on Hetzner Cloud in Germany. The reasons: excellent price-performance ratio, data centers in Germany (GDPR-compliant), fast connectivity in Europe, and a straightforward API. For European SaaS projects, there is hardly a better choice. Data stays in Germany, latency is low, and costs are a fraction of the hyperscalers.

AWS — For Client Projects

I know AWS from over a decade of client projects. EC2, ECS, Lambda, DynamoDB, S3, CloudFront, SQS/SNS — I have worked with all of it. When clients have or need AWS infrastructure, I am at home. But for my own products, I deliberately choose Hetzner: less complexity, better prices, data in Europe.

Azure — When the Client Wants It

I use Azure when clients have a Microsoft-centric infrastructure. That happens regularly, especially in Austria where Microsoft is strongly represented in the enterprise space.

Development Environments

Atoo Studio — My Own Creation

Of course I use Atoo Studio. Not just because I built it, but because it is optimized for my workflow. When I work with AI agents — and I do daily — the integrated environment of editor, terminal, browser preview, and agent interface is exactly what I need.

VS Code — The Classic

For quick edits, configuration work, or when I am working offline, I still reach for VS Code. It is fast, reliable, and the plugin ecosystem is unmatched. My most important extensions: ESLint, Prettier, GitLens, and Docker.

AI Tools

Claude — My Primary Partner

Claude from Anthropic is my primary AI assistant. The combination of strong reasoning, large context window, and reliable code output makes it the best tool on the market for me. I use Claude both through the API in Atoo Studio and through Claude Code in the terminal.

GPT-4 — The Second Opinion

I use OpenAI’s GPT-4 regularly as a second opinion. When Claude gets stuck on a task or I want an alternative perspective, I ask GPT-4. Each model has its strengths, and the combination of both often gives me the best results.

Infrastructure and DevOps

Docker — Everywhere

Docker is the foundation of almost all my deployments. Every service I build runs in a container. The reproducibility and portability Docker provides is irreplaceable.

Terraform — Infrastructure as Code

Terraform is my tool for Infrastructure as Code. I have also tried CDK and Pulumi, but Terraform remains my favorite. The declarative syntax, the state management approach, and the broad provider support are compelling. My Hetzner and client infrastructure is defined in Terraform.

GitHub Actions — CI/CD

For CI/CD, I use GitHub Actions. Not because it is the most powerful tool — that would probably be Jenkins or GitLab CI — but because the integration with GitHub is seamless and I keep all my repositories there anyway. For simple to medium-complexity pipelines, it is perfect.

Frontend

Astro — For Static Sites

Astro is my first choice for static websites and content-driven pages. This very website runs on Astro. The island architecture, the performance, and the flexibility in choosing a UI framework are compelling.

React — When Things Get Dynamic

For complex, interactive applications, I reach for React. Yes, there are alternatives — Svelte, Vue, Solid. But React has the largest ecosystem, most developers know it, and for my purposes it is more than sufficient. I use it together with Next.js or as part of an Astro site.

Databases

PostgreSQL — The Do-It-All Database

PostgreSQL is my default database for everything relational. And honestly, for most applications PostgreSQL is all you need. JSON support, full-text search, extensibility — this database can do practically anything.

DynamoDB — For Specific Use Cases

I use DynamoDB when I need extremely fast read access with predictable latency or when the data model fits key-value access patterns well. It is not a replacement for PostgreSQL but a complement for specific requirements.

Summary

My 2026 stack is pragmatic, not dogmatic. I choose tools that make me productive, not the ones that happen to be trending. The biggest change compared to previous years are the AI tools — they have gone from a novelty to a central part of my workflow. And I expect this trend to intensify in the coming years.