AI should work for you, not against you. Your data belongs to you, not corporations. You should choose your tools, not be locked into one company's ecosystem.

A Major exists because the status quo is broken.

We are a Singapore-based software company. We build agent-native products, starting with Ryu, our flagship orchestration runtime. And we run A Major Agency, which builds custom software for teams who want it done right.

The world doesn't need another AI model

It needs a foundation that makes them all work together.

Every week, a new model drops. A new wrapper launches. A new "agent" promises to change everything. And yet, real AI adoption at scale remains stubbornly low.

Two problems block the path:

1. Security kills adoption

Enterprises can't afford to funnel sensitive data into external systems they don't control. Customer records, proprietary code, financial data, strategic plans, none of it should float around in someone else's cloud.

Every AI tool on the market demands the same devil's bargain: send us your data, and we'll make you productive. Trust us.

IT and security teams know better. They have spent decades building data protection policies. They are not going to throw it all away because a chatbot is impressive.

2. "Agents" that aren't actually agents

Everyone talks about AI agents. But what we have are chatbots with API access, reactive, fragile, and dependent on constant supervision.

Real agents should:

This is why A Major exists, and why Ryu is our flagship.

What we believe

1. Privacy is a right, not a feature

Your screen observations, audio transcriptions, and conversations belong to you. Period.

By default, everything stays on your device. We don't scan it. We don't train on it. We don't sell it. For local processing, we don't even see it.

When you delete something, it's gone. Not archived. Not anonymised. Not kept "for quality purposes." Gone.

Privacy shouldn't cost extra. It should be the default, in every product we ship.

2. Orchestration beats reinvention

The world doesn't need another AI model. It needs infrastructure that makes them all work together.

Every AI tool today forces you into one ecosystem. One model. One vendor. One way of working. But the best agent engine for coding isn't the best for research, and the best model today won't be the best tomorrow.

Ryu wraps any agent engine (Claude, GPT, Gemini, local models) with a unified layer for security, routing, memory, and tools. Pick your engine. Swap it anytime. Ryu handles everything else.

The ultimate unfair advantage isn't a better model. It is the layer that makes every model better.

3. You deserve choice

No single AI model is best at everything. You shouldn't have to pick one and hope for the best.

Switch mid-conversation. Try new models as they launch. Never get locked in.

4. Transparency builds trust

You deserve to know where your data goes, who processes it, how it is used, and when it is deleted.

No fine print. No hidden clauses. No "trust us."

We publish our subprocessors. We explain our architecture. We show our work. When we change something, we tell you.

Trust is earned through transparency, not promises.

5. Local-first is non-negotiable

The cloud is convenient, but convenience shouldn't require surrendering your data.

Local-first means your data lives on your device. Self-hostable means you control the infrastructure. Open means no vendor lock-in, ever.

Built in Rust, not Electron. Single binaries that run on your desktop, your server, or a Raspberry Pi. Your AI infrastructure should never slow you down.

6. Open beats closed

Walled gardens might feel safe, but they are still prisons.

We integrate with every major AI provider. We support MCP servers, sidecars, plugins, and extensions. We build on open standards.

Technology should connect things, not separate them.

7. Reliability beats speed

AI has made it faster than ever to produce software. But the ability to build quickly was never the hard part. The hard part has always been building something that works, keeps working, and earns trust.

Around 41% of all code in 2025 was AI-generated. Change failure rates are up 30%. Technical debt increases up to 41% after AI adoption. The industry is shipping faster and breaking more. They call it "vibe coding": letting AI generate code based on loose prompts, accepting the output with minimal review, and hoping it works. That isn't engineering. That is hoping.

We treat AI as an amplifier, not a replacement. Every line of code gets reviewed. Every system gets tested. We invest in the foundations: process, testing, documentation. The "boring" stuff that determines whether AI helps or hurts.

In a world full of fast, fragile software, reliability is the competitive advantage. And right now, just having something that works consistently is already enough to be ahead.

8. The best AI disappears

The endgame isn't a better chatbot. It is AI that becomes like electricity: invisible, ambient, woven into every device and surface you touch. You won't open an app to use it. It will just be there, running in the background, anticipating what you need before you ask.

This is ubiquitous computing. Mark Weiser described it in 1991: technology so seamlessly integrated into everyday life that you forget it is there. You only notice it when it stops working. Like Wi-Fi. Like running water.

That is the world we are building toward. Local-first architecture, edge computing, always-on agents, multi-device sync. Not because those are trendy buzzwords, but because ubiquity requires infrastructure that works everywhere, runs on anything, and never asks you to hand over your data to function.

Every product we ship is a step toward that future.

9. Users over shareholders

We are building for the people who use our products, not the people who might buy the company someday.

No dark patterns. No surprise price hikes. No selling your data to stay profitable. No pivoting to whatever VCs want.

The moment we optimise for investors over users, we have failed.

Our commitments

We will never:

We will always:

Hold us accountable

This manifesto isn't marketing copy. It is our commitment.

As we grow, we will face pressure to compromise. To add tracking. To store data "just in case." To make our products "more profitable" at your expense.

We won't.

If we ever violate these principles, call us out. We commit to admitting when we are wrong, fixing problems quickly, and being transparent about what happened.

Your trust is earned, not owed. We take that seriously.

A Major. The foundation for AI agents. A software company for the agent era. We build agent-native products, and we help teams ship theirs. Not another model. Not another wrapper. Not another walled garden. Software that just works. Experts you can talk to.