L3 of 4 · How the software those teams build is designed
The Systems‑First Manifesto
Written by Vincent van Gameren
In a world where AI can generate implementation faster than humans ever could, the primary risk is no longer how quickly we write code, but how clearly we define problems, how coherently systems fit together, and how safely they evolve over time.
We are uncovering better ways to build by designing clear systems, validating outcomes early, and using AI to accelerate learning — while keeping decision‑making and accountability human.
Values
Values
While there is value in the items on the right, we value the items on the left more.
SystemsoverSolutions
Solving the right problem matters more than constructing a bespoke answer.
We build systems that address real constraints and outcomes — not isolated implementations that optimize locally and fail globally.
ClarityoverSpeed
AI can move fast. Ambiguity moves faster in the wrong direction.
We prioritize shared understanding, explicit assumptions, and visible trade‑offs before scaling execution.
StandardsoverCustom Plumbing
Boring, well‑understood systems scale.
We default to proven, industry‑standard building blocks for cross‑cutting concerns, and diverge only when the system itself demands it.
Human AccountabilityoverAlgorithmic Output
AI drafts, explores, and accelerates.
Humans decide, commit, and own the consequences — especially when systems affect customers, partners, and businesses.
Principles
12 principles behind the values
- 01
Systems shape behavior — whether we design them or not
Friction, workarounds, and failure modes are system outcomes, not user mistakes.
- 02
AI amplifies system quality, both good and bad
Clear boundaries, standard interfaces, and stable contracts accelerate AI effectiveness.
Ambiguity scales confusion.
- 03
If we cannot articulate the system, we are not ready to build it
Vague intent produces confident, incorrect software — faster than ever before.
- 04
Prefer problem clarity over solution cleverness
Premature solutioning hides real uncertainty and creates systems that are expensive to undo.
- 05
Default to composition; justify invention
Cross‑cutting capabilities (identity, authorization, notifications, observability) should be adopted and composed unless they are core to differentiation.
Custom systems require an explicit reason.
- 06
Industry standards are force multipliers for AI
AI implements and integrates established patterns more reliably than bespoke abstractions.
Standards reduce friction — for humans and machines alike.
- 07
Differentiation belongs where it creates customer value
We invest custom engineering effort where it materially improves the experience or the business — not where it recreates solved problems.
- 08
Validation beats debate
When experiments are cheap, opinions are expensive.
We validate the riskiest assumptions with evidence before turning beliefs into architecture.
- 09
Validation gates remain human‑owned
AI accelerates learning, not responsibility.
Decisions about viability, feasibility, risk, and safety belong to people.
- 10
Design for recovery, not perfection
Failures are inevitable.
Systems earn trust by making failure visible, contained, and recoverable.
- 11
Optimize for end‑to‑end flow, not local output
A fast team inside a slow system is still slow.
We optimize the whole, not the parts.
- 12
Change is safest when systems are explicit
Clear ownership, clear contracts, and clear boundaries make evolution possible without fear.