Niche projects — no project is too small

Niche projects — no project is too small

I work as an agency employee on large Magento and related stacks. I also see a future prospect in careful niche custom work: small web + mobile applications that used to die on budget, and that AI-assisted delivery can now make reachable — for freelancers, small agencies, NGOs and organisations that need a tool that fits them, not a SaaS that fits nobody.

Why this belongs near big monolithic projects (like Magento) — and near microservices-orchestrated stacks

Agency employment and niche custom work are not enemies. The first pays the bills and sharpens craft on complex platforms — whether that is a large Magento / Mage-OS monolith or a microservices estate you orchestrate day to day. The second is a hedge and a vocation: AI is compressing the cost of building and maintaining software, which also means some “employee as coding machine” roles are at risk over a few years. I have written about that labour-and-ethics framing in Reading Obsolete and related notes. The constructive answer is not despair — it is to use the productivity boost for projects that were previously unthinkable because the budget could never carry a classic custom build or years of expensive maintenance.

Ready-to-go stacks change the unit of time. Take Umbraculum (docs) as an illustrative example only — not an endorsement and not “the one stack.” Combined with a React Native mobile surface (Expo, shared TypeScript UI), you can stand up a custom mobile-first native application with web support in hours: shared logic, forkable licences, typed contracts, and humans who still own the judgment when models help write the code. That is a different scale from both the Magento programme and the microservices programme — a thin, owned tool that sits beside them when the organisation needs something SaaS will not customise.

What “niche” means here

  • Small applications that improve how an NGO, co-op, club or small company actually works day to day
  • Web + mobile when that is the real surface (not a brochure site pretending to be a product)
  • Delivery measured in weeks or a few months, not multi-year programmes
  • Maintenance that stays affordable because the stack is forkable, typed, tested and owned by humans who can still read it

Choosing a stack carefully (not pitching one product)

For customer projects that want web + mobile from shared logic, one illustrative example I use in conversation is Umbraculum (brochure, docs) — not because I run it as my own daily product, and not as an endorsement or preferred vendor pitch. I first encountered it in early Cursor chats; its licences and project posture (open, forkable, serious about governance) make it a sensible reference for niche builds. There are many other options. What matters is choosing with caution:

  • Forkability — can you leave without begging a vendor?
  • Licence — AGPL / copyleft / OSI-clear terms you can explain to a board
  • Right tools and languages — TypeScript/PHP/etc. your maintainers can hire for
  • React Native + Tamagui when mobile is real — a huge win if you target native iOS/Android with shared UI primitives: you build once and, so to say, get web + native done on the first pass instead of maintaining two forever-divergent apps (see Mobile apps)
  • Ethics — privacy, labour, and “who owns the judgment when AI helps write the code”

More detail — and concrete application sketches — in the note Niche projects as future prospect.

Visit the React Native page to see how React Native + Tamagui fits this picture — and how a ready-to-go stack (shared TypeScript, forkable licences, humans still owning the judgment) unlocks the potential of niche projects that budget once killed.

← All services · AI & bots · Contact