Manifesto

OmO is opinionated. The features exist because of a small set of principles, and most product decisions follow from them. This page is the principles, not the feature list.

Human intervention is a failure signal

Autonomous driving: if a human has to take the wheel, that is not a feature, it is the system failing. The car could not handle the situation on its own.

Coding is the same. When you find yourself fixing the AI's half-finished code, manually correcting obvious mistakes, guiding the agent step-by-step, or repeatedly clarifying the same requirements — that is not collaboration. That is the AI failing to do its job.

OmO is designed so that intervention is the exception, not the loop.

Indistinguishable code

Code written by the agent should be indistinguishable from code written by a senior engineer. Not "AI-generated code that needs cleanup." Not "a good starting point." The actual, final, production-ready code.

In practice:

• Existing codebase patterns followed exactly.
• Error handling done without being asked.
• Tests that test the right thing.
• No AI slop — no over-engineering, no scope creep, no decorative abstraction.
• Comments only when they add value.

If you can tell whether a commit was made by a human or an agent, the agent has failed.

Token cost vs productivity

Higher token usage is fine when it increases productivity. Parallel research, end-to-end completion without intervention, thorough verification, and accumulated knowledge across tasks are worth the spend.

Unnecessary waste isn't. The system optimizes for:

• Cheaper models (Haiku, Flash, GPT-5.4 Mini) on the utility lane.
• No redundant exploration when context is already sufficient.
• Caching learnings across sessions.
• Stopping when "enough" is reached.

Token efficiency matters, but not at the cost of work quality or human cognitive load.

Minimize human cognitive load

The human's job is to say what they want. Everything else is the agent's job.

Two paths to the same result:

• Prometheus mode — say "I want to add authentication." Prometheus researches the codebase, asks clarifying questions grounded in actual findings, surfaces edge cases you hadn't considered, and writes a plan. You provide intent; the agent provides structure.
• Ultrawork mode — say ulw add authentication. The agent figures out the right approach, researches best practices, follows conventions, and keeps going until done. You provide intent; the agent handles everything.

The human expresses what they want, not how it gets done.

Predictable, continuous, delegatable

The ideal agent works like a compiler: markdown in, working code out.

• Predictable. Same codebase patterns, same requirements, same constraints → consistent output. Not random, not surprising, not "creative" in ways nobody asked for.
• Continuous. Work survives interruptions. Session crashes resume with /start-work. Multi-day projects preserve context. The agent maintains state; you do not.
• Delegatable. Clear acceptance criteria, verified independently. Self-correcting on failure. Escalation only when truly needed. Complete work, not "mostly done."

The core loop

Human Intent  →  Agent Execution  →  Verified Result
      ↑                                      ↓
      └──────────────── Minimum ─────────────┘
        (intervention only on true failure)

The OmO surface area maps onto this loop:

What this means in practice

You should be able to:

1. Describe what you want at any level of detail.
2. Let the agent interview you when scope is unclear.
3. Confirm the plan, or skip straight to ulw.
4. Walk away.
5. Come back to completed, verified, production-ready work.

If you cannot do this, something in the system needs to improve.

The future being built

A world where human developers focus on what to build, not on how to get AI to build it; where code quality is independent of who wrote it; where complex projects are as approachable as simple ones, just slower; and where "prompt engineering" becomes as obsolete as "compiler debugging."

The agent should be invisible — not hidden, but reliable enough that nobody has to think about it. Like electricity. You flip the switch, the light comes on. You don't think about the power grid.

Source Notes

Aligned with upstream docs/manifesto.md. The product reality check (name, domain, who maintains it) lives upstream; this site keeps the principles a doc rather than a status page.