ABOUT /SEREIN

Serein Team is an independent systems team.
We believe good software should work like precision instruments — accurate, reliable, and dependable.

SECTION 01 / 05

Who We Are

Not for the many, but for those who refuse to compromise.

Serein Team was founded in 2025, rooted in Hong Kong. We are an independent systems team focused on systems-level software and programming languages.

We don't chase user numbers or industry trends, and we don't rely on marketing gimmicks. We do one thing: make every tool stand up to scrutiny.

"Serein" means clear-sky rain — rain falling through sunlight. We chose the name because good engineering often goes unseen, yet it is always there, quietly doing its work.

Our flagship product is Aero — a systems programming language built on LLVM. Write like Python, run like C++, memory-safe at compile time, no GC.

What We Believe

Four principles.
One standard.

Precision FirstCORRECT SEMANTICS

Semantics must be precise and behavior must be predictable. Whatever the compiler can catch, it catches — and when something is wrong, it says so rather than silently returning a wrong answer.

Native FirstNATIVE BY DEFAULT

Where system APIs can be called directly, we don't wrap a framework. Where we can compile to native machine code, we don't lean on an interpreter. One abstraction less is one overhead less — performance and correctness are non-negotiable.

Restraint FirstPURPOSEFUL INTERFACE

No superfluous features, no meaningless decoration. Language features stay few and orthogonal; every element in an interface exists for a reason. Less is more — not out of laziness, but out of respect.

Ready FirstSHIP WHEN READY

No rushing to release. We verify on more platforms and test in more edge cases. A version number can wait; crashes and regressions cannot.

How We Ship

Human-AI collaboration. Numbers do the talking.

AI-AssistedAI · ~55%

Scaffolding and boilerplate, first-draft implementations, edge-case generation, performance baselines.

Team-LedTeam · ~45%

Architecture and semantics decisions, code review, systems-level debugging, verification and release gates.

ReproducibleReproducible

aero-csv processes 2M CSV rows (filter + aggregate) in 54.4 ms — 10.4× faster than NumPy. All three implementations return identical results (count=950396 / total=2496544296331). Commands and scripts are public.

How the Team Operates

We train AI — it ships, we guard.

AI-drivenTrained Models as Tools

We train models into productivity: scaffolding, boilerplate, edge cases, first-draft implementations — AI ships, we judge, no scribe labor. Code contribution ~55% AI / ~45% team.

Human GatekeepingThree-Layer Audit

AI output never lands in the repo unchecked. Every contribution goes through vulnerability scanning, hallucination detection (deepfake-style fake-code rejection), and performance benchmark — then team review. Better slow than broken.

GitHub AI Stewardship~50% Repo Decisions by AI

We are training AI to autonomously steward the Serein GitHub organization — triaging issues, reviewing bot MRs, maintaining CI and repo health. ~50% of repo maintenance decisions are executed by AI independently; humans review every trail. This is our technical runway and the long-term capability we are building.

GET IN TOUCH