Manifesto · v2026.05

How we think.
What we promise.

A small group of senior engineers needs few rules but very firm ones. This is the document that holds them together — our internal contract, public because anyone working with us should be able to read it first.

Document · kytheron / manifestoLast updated · 2026.05Status · v1 (living)
/01 · Why we exist

Most companies don't need more strategy.

We believe most companies don't need more frameworks, more decks, or more consultants explaining what to do. They need a small group of technical people who walk in, read the code, and make things work.

KytheronLabs was founded in 2026 in Naples as an independent group of senior engineers. We work where engineering stops being an option and becomes the only lever that actually moves the business: cloud migrations, event-driven architectures, AI-native systems serving real production.

We are not an agency. We are not a software house. We are an independent collective of senior engineers — whoever sells a project is the same person who executes it.

/02 · The principles

Six rules we don't negotiate.

These are the criteria by which we decide whether to take on a client, how to engage, when to say no, and what to leave behind at the end of an engagement.

/01 · ship

Ship, don't pitch.

Every sprint's deliverable is code in production — not a PDF, not a slide, not a Notion page. If nothing went live by week's end, we wasted the client's time.

/02 · senior

Only people who can do the work.

No pyramids. Every person you see at kickoff is the one writing the code. We don't sell one profile and put you with someone else — it's a practice we openly despise.

/03 · boring

Boring tech, smart application.

Postgres before Cassandra. Predictable before novel. Tech novelty has a cost — clients pay for outcomes, not for your enthusiasm for the latest framework.

/04 · document

Document the why.

For every non-trivial decision we write an ADR — Architecture Decision Record. The "why" outlives us. The next maintainer shouldn't have to do archaeology to understand the choices.

/05 · transfer

Leave it better.

When we leave, the system is simpler than when we walked in. Documented, tested, maintainable. If the internal team can't keep it running without us, we failed.

/06 · honest

Brutally honest, never cruel.

We say "I don't know" when we don't know. We say "this choice is wrong" when it is. We say "we're not the right vendor" when we're not. Courtesy + truth, never one at the expense of the other.

/03 · Conscious AI

AI is a primitive, not a solution.

We treat Large Language Models the way we treat any other system component: measurable, reproducible, observable, debuggable. No magic. No hype. No prompts hastily engineered that accidentally become critical infrastructure.

/operating rule

If an AI agent in production can't be monitored like a microservice — with tracing, structured logs, alerts, SLOs and rollback — then it isn't in production. It's a demo that happens to live on a server.

/ai · 01

AI-native, not AI-washed.

We build as if LLMs, embeddings and agents were default primitives. But we don't add "AI" to a problem that doesn't need it just to appease a stakeholder.

/ai · 02

The prompt is code.

Versioned in git, tested with evals, subject to code review. An unwritten prompt is a waiting bug — an untested prompt is a waiting incident.

/ai · 03

When AI isn't the answer, we say so.

Sometimes a well-written SQL query beats a RAG. Sometimes a cron job solves what an agent complicates. Technical rigor comes before alignment to market narrative.

/ai · 04

Client data, client sovereignty.

No silent fine-tuning. No data leaving the perimeter without explicit authorization. GDPR and privacy aren't compliance checkboxes — they're engineering.

/04 · Agile as a means

Agile is a means, not an identity.

We've read the original 2001 Agile Manifesto — the 68-word one, before it became a certification industry. That one, yes. Everything stacked on top is optional: frameworks, rituals, ceremonies, scaling models. Tools, not religion.

/lean · 01

Short cycles, reversible decisions.

We ship small and often. Big decisions are made only when they have real effect on the system — not early, not late, not "just to be safe".

/lean · 02

Feedback from the system, not from meetings.

Truth lives in production: metrics, logs, real errors, real users. Stand-up = quick sync point, not a ritual. Retro = a moment to change something, not to fill a template.

/lean · 03

Public roadmap, brutally honest retro.

The client sees where we're going, always. When we screw up, we write it in black and white along with the cause and the fix. Postmortem is an engineering practice, not a blame exercise.

/lean · 04

No process for process's sake.

If a ritual doesn't produce decisions or code, we cut it. If a tool creates more overhead than value, gone. Bureaucracy is the way technical teams die.

/05 · The promise

What we guarantee when you sign.

/contract

No account managers. No middlemen. No "team we'll tell you about later." When you sign an engagement, you know the names of the people who will work on your system. They're the same names you saw at the first meeting.

/promise · 01

Transparent pricing.

No hidden rate card, no discount theater. What you pay is written down, and it covers the work — not the organizational layers.

/promise · 02

Code is the client's property.

Repository, infrastructure, knowledge: all the client's, from day one. No hidden lock-in, no "consulting tax" on the way out.

/promise · 03

We say no.

If the project isn't for us — by scope, stack, timing, or ethics — we say so immediately. Three right engagements are worth more than ten wrong ones.

/promise · 04

Success = client autonomy.

Our goal is the moment we're no longer needed. We don't sell dependency. We sell capability that stays inside the company.

KytheronLabs Collective
Naples, May 2026 · v1 · living document
/start

All clear? Let's talk.

Book a discovery call