Skip to main content
Tech radar

Our tech radar: strong opinions, loosely held.

What we use by default, what we're experimenting with, what we're watching, and what we advise against for new projects. Updated as we learn.

The radar

Adopt: our default. Trial: use on suitable projects. Assess: worth exploring. Hold: avoid for new work.

Adopt

  • PostgreSQL

    The answer to most 'which database?' questions.

  • TypeScript

    Types are documentation that the compiler checks.

  • Next.js App Router

    Server-rendered by default, interactive where needed.

  • FastAPI

    Typed Python APIs with free OpenAPI docs.

  • Terraform / OpenTofu

    Infrastructure you can review in a pull request.

  • Structured LLM outputs

    Schemas over string parsing, always.

Trial

  • pgvector

    Vector search without another database to run.

  • OpenTelemetry

    One standard for traces, metrics, and logs.

  • Playwright

    End-to-end tests that are actually stable.

  • Edge-deployed functions

    Great for light logic close to users.

Assess

  • Agent frameworks

    Moving fast; we keep agents small and custom for now.

  • WebGPU

    Promising for heavy 3D and compute in the browser.

  • Local-first sync engines

    Could simplify offline apps significantly.

Hold

  • Patching Odoo core

    Upgrade pain with compound interest.

  • Microservices by default

    Start with a well-structured monolith.

  • Kubernetes for one app

    Lots of machinery for a small job.

  • Parsing free-form LLM text

    It works until the model updates.

How we use the radar

The radar is how we keep technology choices deliberate. When we start a project, items in Adopt are the default unless there's a good reason otherwise. Trial items are used where they clearly fit and someone experienced can support them. Assess items are being explored in prototypes and internal work. Hold means we advise against starting new work with it, though existing systems built on it may be perfectly fine.

Every item has a one-line reason because the reason matters more than the label. A tool can be excellent and still sit on hold for a specific kind of project if there's a simpler or safer option.

How items move

Things move when our experience changes. A trial item moves to Adopt after it has served several projects well and the team can support it confidently. An adopted item can slide back if maintenance slows, licensing changes, or a clearly better option appears. We would rather update the radar honestly than defend yesterday's choice.

How we evaluate a new technology

  • Does it solve a problem we actually have?
  • Is it mature, maintained, and well documented?
  • Can we hire for it, or teach it, in reasonable time?
  • What does it cost to run and to leave?
  • Does it fit our security and accessibility standards?

Using it for your own decisions

Treat the radar as one input, not a rulebook. Your team's skills, your existing systems, and your constraints may point to different choices, and that's fine. If you're evaluating a proposal that leans on something we've placed on hold, ask the proposer why; there may be a good reason, and the answer will tell you a lot.

The rings in practice

RingWhat it means for a new project
AdoptOur default choice; we use it unless there's a specific reason not to
TrialUsed where it clearly fits, with an experienced engineer involved
AssessExplored in prototypes and internal tools before client work
HoldNot recommended for new work; existing systems may still be fine

Questions about a choice?

If you'd like to know why something sits where it does, or how it applies to your situation, ask us. We're happy to explain our reasoning, including when your context points somewhere else.

What the radar isn't

It isn't a ranking of the best tools in the world, and it isn't sponsored. It reflects what has worked for the kinds of systems we build and the teams we hand them to.

Why 'Hold' doesn't mean 'bad'

A technology on Hold may be excellent in the right hands. It sits there because, for the kind of projects we do, the cost of adopting or maintaining it outweighs the benefit, or because a simpler option does the job.

If you already run something on our Hold list successfully, keep running it. The radar is about what we'd choose for new work, not a judgement on systems that already serve you well.

What each ring means for a project

RingWhat we'll do on a new project
AdoptUse it by default unless there's a reason not to
TrialUse it where it fits, with a fallback plan
AssessPrototype or research only, not production yet
HoldAvoid for new work; maintain where it already runs

How a technology earns a place

New tools join the radar after we've used them on real work, not after a conference talk. We look at maintenance health, documentation, how easy it is to hire for, how it fails, and whether it makes a client's system simpler or just more interesting.

Frequently asked questions

How often is the radar updated?
Whenever our experience changes a recommendation, typically after a technology has been tried on real projects rather than on a fixed schedule.
Can we use the radar for our own decisions?
Please do. Treat it as one experienced opinion to compare with your own constraints, not as a rulebook.