The manifesto.
This is the document that defines what Omnian is, why it exists, and what it will never be. If you need to understand the product in five minutes, read to the end.
What Omnian is
Omnian is a platform for AI-assisted intellectual work — built to last longer than one chat session.
Most AI tools today treat every conversation as disposable. You open, talk, close, lose. When you come back, you explain it all again. Context evaporates, work fragments, and the AI — however brilliant in the present minute — knows nothing about you the next.
Omnian flips that. The conversation is eternal. The session is a hidden technical detail. The project remembers itself. The AI arrives at work knowing where things were left.
In one sentence
Omnian is the cockpit where professionals run continuous work with AI, in an environment that remembers context, preserves understanding, and steps out of the way so you can produce.
The non-negotiable principles
There are five. Every product, design, or architecture decision passes through them. When a principle conflicts with a feature, the principle wins.
Principle 01
Continuity over novelty
The person should not feel they've started a new session. They should not have to think "let me remind the AI of what we talked about". They should not have to copy summaries, paste contexts, repeat explanations. Continuity is the system's responsibility, not the user's.
Internally, sessions exist — because LLMs have a finite context window. Externally, sessions disappear. They surface only when the user explicitly asks: "what did I say about X last week?".
Principle 02
Project knowledge belongs to the person, not to the AI
The AI does not decide what to remember. It has no memory of its own about you. What exists is a structured layer of project knowledge — rules, decisions, glossary — that belongs to the person, is auditable by them, and is editable.
The AI consults. The person decides.
This protects against the two evils of naive conversational memory: the echo chamber (the AI reinforces your wrong interpretations as if they were facts) and opacity (the person does not know what the AI "thinks it knows" about them).
Principle 03
Recognition before learning
When someone from field X opens Omnian for the first time, the brain has to go straight to "oh, this is like [thing I already use], just with AI alongside". In two seconds. No manual.
That's why Omnian presents itself as a vertical, not as a generic tool. The developer opens it and sees something that looks like a Cursor with conversational AI. The researcher will open it and see something that looks like Zotero+Obsidian with AI.
Same engine underneath. Specific vocabulary, icons, and structure on top.
Principle 04
Calm tone, always
The AI industry has become addicted to alarms. "Your session is getting long." "The model may lose details." "Warning: reduced context." All shouted, in red or yellow, with exclamation marks.
Omnian doesn't shout. When it needs to signal something (context saturation, session change, usage limit), the tone is one of continuity. The copy describes what is preserved, not what is lost. The color is olive, welcoming, not red. The button says "Continue in a new session", not "Reset".
The person needs to feel they're in competent hands, not that the tool is about to fail.
Principle 05
Generic core, specific experience
Behind every mask there is a technical core that doesn't know there is such a thing as "code", "thesis", or "contract". It knows there are artifacts, knowledge, sessions, and continuous work.
Every vertical (developer focus today, academia later, others ahead) is a declarative mask over that core: vocabulary, icons, artifact types, modes of operation. Zero new code to create a mask. It's configuration.
This is not architectural luxury. It's what lets Omnian be born focused on dev and grow into other worlds without becoming three parallel products.
What Omnian is NOT
As important as defining what we are is defining what we are not. Each item below is a temptation we'll resist.
- Omnian is not a chatbot. Chat is the interface, but the product is continuous work. The difference is everything: continuity, knowledge, files, actions.
- Omnian is not an IDE. Even in the dev vertical, we don't compete with Cursor, VSCode, or Copilot on the side of "code editor with AI". We compete on the side of "work environment that remembers you". The editor is means, not end.
- Omnian is not an opinion-less platform. It's not Notion-blank, it's not a blank page. Every mask has strong opinions about how work in that field should flow. Extreme customization comes later and is the exception, not the rule.
- Omnian is not a prompt engineering tool. The person shouldn't have to think about prompts, contexts, tokens, models. All of that is the system's problem. The user thinks about the work.
- Omnian is not multi-AI mixed together. Internally we may route to different models depending on the mode (Fast / Standard / Deep). But the user never chooses the model. The abstraction is over modes, not providers.
- Omnian is not free by principle. AI has real cost. We charge for usage honestly and visibly, with no tricks. No degraded fallbacks, no "lite" versions that mislead.
Who the user is
Omnian is born focused on technology professionals — developers, software engineers, architects, technical leaders. People who:
- Work on long, complex projects spanning more than a week
- Already use AI in some way today, but get frustrated by the lack of continuity
- Value tools that step out of the way (they don't want another app to manage)
- Have their own budget or influence over team budget
Further ahead, the same platform serves companies (governance, roles, teams) and academia (research, academic writing, learning). The architecture allows it. The initial focus does not.
What the person feels when using Omnian
This is the most important success metric. More than MRR, more than NPS, more than retention. If the person feels this, the rest follows.
- "The tool remembers me." — not in an invasive sense, but in the sense of professional continuity. The way a colleague would remember.
- "I don't have to re-explain." — projects have rules, decisions, context. Omnian already knows.
- "I can trust what it does." — every action is visible, auditable, editable. No black magic.
- "It steps out of the way when I need to focus." — calm UI, no alarms, no pop-ups, no distractions.
- "I'm doing real work here." — not a chatbot toy. It's production.
The bet
The whole architecture, the whole product, the whole effort, is betting on a premise:
The next generation of AI tools won't win by having the smartest model. It will win by understanding the continuity of human work.
Models will commoditize. Cap rates will fall. APIs will standardize. What will set winners apart is who architects the work around the AI — not the AI around the chat.
Omnian is the bet that this can be done. And that it's worth doing now.
The conversation is eternal.
The project remembers.
Omnian begins where the chat ends.