Lyra by Roxabi — Messaging Refinement v2

Taglines & Refined Positioning

The "Extend" direction. Bidirectionality as the brand's core tension: you extend Lyra, Lyra extends you.

The Extend Direction — 17 Options
Four groups. Each explores a different angle on the bidirectional extend concept. Top picks are highlighted.
A — Bidirectional Extend

The duality: the tool shapes the maker, the maker shapes the tool.

A-02
The more you shape it, the more it shapes you.
Longer — better for subheadings or product page intros than a hero tagline.
A-03
Built to be rebuilt. Runs to keep running.
Two halves mirror the bidirectional value: extensibility + always-on.
A-04
Your additions. Its compounding.
Finance-adjacent. Resonates with Yuki's efficiency mindset. Short and precise.
B — Builder / Craft

For the tinkerer who wants to OWN their stack. Speaks to Yuki directly.

B-02
Fork it on day one.
Dev vernacular. Signals the architecture is readable and meant to be touched.
B-03
Assemble your intelligence engine.
Category-forward. Positions Lyra as a platform, not a fixed product.
B-04
Built for the ones who read the source.
Strong in-group signal. Yuki reads source code. This is a call to her.
B-05
Minimal core. No ceiling.
Echoes the architecture: ~300-line hub, infinite extension surface.
C — Intelligence Amplification

You + AI > either alone. The "intelligence engine" concept — not replacing you, augmenting you.

C-02
The engine underneath your thinking.
Subordinate positioning — Lyra is infrastructure, not the hero. Technically honest.
C-03
Intelligence that runs on your terms.
Sovereignty implicit. "Your terms" does more work than "self-hosted".
C-04
Think further. Ship faster. Own everything.
Three-beat rhythm. Works as a hero or section header. Action-oriented.
D — Quiet Confidence

Understated and powerful. For the dev who will run from marketing fluff.

D-01
Always running. Always yours.
Two promises. Availability + sovereignty. Clean and memorable.
D-03
It runs. You extend. It learns.
Three clauses, each adding a layer of value. Prose version of the architecture.
D-04
No subscription. No ceiling. Just yours.
Anti-SaaS positioning. Resonates with privacy-first and self-sovereignty motivations.
The One Statement
Combines category, target, lead angle (intelligence), and extensibility as the emotional hook. Relationship framing reserved for community channels.
Canonical Positioning Statement

For developers who self-host and build on their own terms,

Lyra is a Personal Intelligence Engine — an always-on, extensible hub that compounds your capabilities across every tool, channel, and workflow you own.

Unlike cloud AI assistants that lock in your data and limit what you can build,
Lyra runs on your hardware, learns your workflows, and grows as you extend it — giving back more than you put in.

Category
Personal Intelligence Engine
Target
Developers / tinkerers who self-host and build
Lead angle
Intelligence — capability, memory, multi-agent
Emotional hook
Extensibility — bidirectional value compounds
Relationship language
"Learns your workflows" — workflow-framed, not surveillance-framed
Differentiator
Runs on your hardware. You extend it, it grows with you.
Usage Guidance

Landing page & README: Lead with the category name "Personal Intelligence Engine" in the hero. Follow immediately with the capability proof: always-on, memory, multi-agent routing, extensible adapters. The positioning statement above can run as a subheading or an About section verbatim.

Conference talks & demos: Open with the problem ("every useful AI tool sends your data somewhere you don't control"), then introduce the category. Let the architecture do the positioning — show the hub, the bus, the pools. The statement is implicit in the demo.

Community channels (early adopters): Only here does "learns your workflows" become the lead. The relationship framing is earned, not announced. Use it in onboarding copy, in-product messages, and persona-level outreach — not on the public homepage.

30 Seconds, Two Audiences
Spoken versions. No jargon in the non-dev version. No fluff in the dev version.
Developer Version — Technical, precise
"Lyra is a self-hosted intelligence engine — a hub that runs 24/7 on your own hardware, connects to any messaging channel, and routes conversations to specialized agents using your own LLMs, memory, and tools.

The core is about 300 lines. Everything else — adapters, agents, skills — is an extension point. You own the stack: no cloud lock-in, no subscription, no data leaving your machines.

What makes it different: it's bidirectional. You build on top of it. It compounds what you can do. Fork it on day one, ship something in an afternoon."
Non-Dev Version — Accessible, intriguing
"Imagine an AI assistant that actually runs on your own computer — not on some company's server — and that you can connect to Telegram, Discord, or any app you already use.

It remembers your context, understands your routines, and gets smarter the more you use it. But unlike ChatGPT or similar tools, nothing leaves your machine, you don't pay a monthly fee, and you can shape it however you want.

We call it a Personal Intelligence Engine — not a product you subscribe to, but a piece of infrastructure you own."
Key Differences by Audience

The dev version leads with architecture signals. "300 lines", "extension point", "fork it" — these are not for everyone. They are deliberate shibboleths. Yuki hears "300 lines" and immediately understands: auditable, lightweight, no magic. Non-devs hear it and move on. That is intentional.

The non-dev version leads with the outcome, not the mechanism. "Runs on your own computer", "remembers your context", "nothing leaves your machine" — these are the same claims, translated. The category name ("Personal Intelligence Engine") appears at the end as a label for what they have already understood, not as an upfront definition they need to decode.

"Why Lyra" — Section Map
The structure and emotional arc of the one-pager. Copy belongs in a separate pass. This is the container.
1
The Problem — The Broken Trade-Off
Emotion: Recognition
What goes here: A tight matrix of the current landscape — cloud AI (your data leaves), frameworks (you build everything), open-source bots (no memory, no channels), visual SaaS tools (hard to extend, no real agent reasoning). Every existing option forces a painful compromise.
  • No names of competitors — describe the archetype, not the company
  • Language mirrors what Yuki has already said to herself
  • End with the gap: "there is no option that gives you all four"
2
The Claim — A Different Category
Emotion: Curiosity
What goes here: Introduce the category "Personal Intelligence Engine". One sentence definition. Then the three properties that define it: runs on your hardware, extends in every direction, compounds over time. Not a feature list — a promise.
  • Category name stands alone — give it a visual moment
  • The three properties map to: Sovereignty, Extensibility, Memory — in that sequence
  • No benchmark claims, no "10x faster" — just the shape of the thing
3
The Architecture — Proof in the Structure
Emotion: Desire
What goes here: A minimal diagram of the hub-and-spoke model. Hub (~300 lines), channels, agent pools, memory, skills. Not a deep-dive — just enough to show that the architecture is clean, legible, and intended to be touched. This is where Yuki leans in.
  • Diagram does more than copy here — use one
  • "Auditable by design" as a callout — every routing decision is traceable
  • End with the extension surface: adapters, agents, skills — all yours to write
4
The Proof — What It Actually Does
Emotion: Conviction
What goes here: Two or three concrete scenarios — not features, but moments. Example: "You paste a client schema into Telegram. Lyra answers. Nothing left your machine." Or: "You wrote a skill on Saturday. On Monday it ran without you touching it." These are outcome-oriented vignettes, not tutorials.
  • Each scenario is one paragraph — tight
  • Scenarios should span the three properties (sovereignty + extensibility + memory)
  • One scenario should be technical; one should feel almost mundane in the best way
5
The Call to Action — Start with the Engine
Emotion: Action
What goes here: One primary action ("Get started" / "Fork the repo" / "Read the docs") and one secondary action for the community channel (Discord/GitHub Discussions). CTA language should match the "builder" register — not "Sign up free" but something closer to "Clone it. Run it. Make it yours."
  • No pricing on the one-pager — Lyra is not a SaaS (at this stage)
  • Brief "what you need" tech requirements — one line each: hardware, OS, knowledge
  • Optional: a quote from an early adopter framed around what they built, not how they feel
Intelligence vs Relationship — When to Use Each
A 4-part analysis: when to use each framing, and how to make Relationship work without triggering surveillance anxiety.
Intelligence Framing
Lead with capability and architecture
  • Landing page hero and above-fold copy
  • README introduction and docs front matter
  • Conference talks, demo scripts, screencasts
  • Technical blog posts and architecture writeups
  • Hacker News / Lobsters / dev.to launch posts
  • Competitive comparisons and positioning tables
Relationship Framing
Lead with workflow and continuity
  • Early adopter outreach (DMs, intro emails)
  • Community messaging (Discord, GitHub Discussions)
  • Onboarding copy and first-run experience
  • Testimonial framing and case study structure
  • Retention messaging and changelog announcements
  • Newsletter and personal update cadences

Intelligence framing is the lead everywhere a stranger reads you first. When Yuki lands on your README or sees your landing page, she is evaluating a tool, not a relationship. She wants to know what it does, how it does it, and whether the architecture is trustworthy. "Personal Intelligence Engine" signals a category claim she can verify. It respects her as an engineer. Intelligence framing is honest, precise, and impossible to misread as surveillance.

Relationship framing earns its place after trust is established. Once Yuki has cloned the repo, run the hub, and built her first skill, the experience starts to feel personal in a way she chose. This is when Relationship language becomes accurate — and only then. In community channels, early adopter DMs, and onboarding copy, "Lyra learns your workflow" is not a marketing claim. It is a description of what actually happens. The key move: frame it as workflow, not identity. "Lyra adapts to your patterns" is grounded. "Lyra knows you" is creepy.

The language that makes Relationship work without the surveillance edge: anchor every claim in what the user chose to do. "You added a memory namespace for your work notes — Lyra uses that context in every subsequent query." Not "Lyra remembers you." The user is the agent in these sentences. Lyra is the engine that serves what the user deliberately configured. Autonomy and consent are in every word choice.

What to avoid in both framings: Anthropomorphism that implies awareness Lyra does not have ("Lyra cares about your privacy"), possessiveness that implies lock-in ("your Lyra"), and performance claims that substitute for proof ("the smartest personal AI"). The brand's tone is understated and technically honest. It trusts the product to speak through the architecture, not through adjectives.

Language Examples — Do / Avoid
Avoid — surveillance framing
"Lyra knows you better every day."
Use — workflow framing
"Lyra adapts to how you actually work."
Avoid — identity-level claim
"An AI that understands who you are."
Use — capability claim
"An intelligence engine that compounds your workflows."
Avoid — passive personalization
"Your AI learns your preferences automatically."
Use — active, consent-first
"You shape the memory. Lyra uses what you give it."
Avoid — anthropomorphism
"Lyra cares about keeping your data private."
Use — architecture-grounded
"Your data never leaves the machine. Not by policy — by design."