The Big Idea in One Sentence
Flow is a platform where anyone can earn money by contributing knowledge and work, powered by a personal AI that represents them and handles the complexity on their behalf.
We are designing something genuinely novel to feel immediately familiar. Most users will not understand Bittensor, may be skeptical of AI agents, and have never used a decentralized marketplace. The UI has to do the heavy lifting of making all three feel obvious & very intuitive.
The User Journey (End to End)
1. Discovery & First Impression
The landing page has one job: answer "what's in it for me?" within 5 seconds. Not "we're building decentralized AI infrastructure" but "earn money doing what you're already good at. Your AI handles the rest."
The hero should show a real task, a real payout, a real person. Social proof from day one matters enormously here, especially for the African market where trust is earned through word of mouth, not just brand.
"Your AI assistant" not "your autonomous agent."
"Earnings" not "token rewards."
"Skills marketplace" not "decentralized workstream."
2. Sign Up
Keep it frictionless. Email or phone number to start, phone number first given the target market. No wallet setup on day one. No crypto knowledge required.
The onboarding asks three simple questions:
What are you good at? (presented as skill tags, not a blank form)
What kind of work are you looking for? (quick/small tasks vs. bigger projects)
How much time do you have per week?
That's it. Jarvis takes those answers and builds the foundation of their profile. The user never sees "you're creating a knowledge base." They see "we're setting up your assistant."
3. Meet Jarvis (The AI Setup)
This is the most critical UX moment in the entire product. If users don't understand or trust Jarvis, nothing else works.
Frame it as: "Meet your work assistant. The more it knows about you, the better work it finds you and the more it can handle on your behalf."
Introduce Jarvis through a short conversation, not a form. It asks about what they've done before, what they're proud of, what they want to earn. It feels like talking to a smart recruiter. Behind the scenes, it's populating their private knowledge base.
Then show them what Jarvis can do for them in plain terms:
"I'll match you to tasks that fit your skills"
"I'll handle back-and-forth with task posters"
"I'll notify you when something needs your decision"
Start Jarvis on low autonomy, it suggests, the user approves. Autonomy increases as trust builds. This is not a settings toggle; it's something that happens naturally as they use the platform.
4. WorkStream (Finding & Doing Work)
The feed is the product here. It needs to feel like a smart inbox, not a job board. Jarvis has already filtered and ranked tasks before the user ever opens the app.
Each task card shows: what the work is, how long it will take, what it pays, and why Jarvis thinks it's a match. One tap to accept, one tap to decline. No lengthy application process.
For first-time users, the first task should feel almost too easy, a small, well-scoped task that builds confidence and gets them their first payout. The apprenticeship model from the doc maps directly here: complex tasks are decomposed and junior contributors get the appropriate slice.
Task completion triggers a simple review flow. Did it meet the brief? Any feedback? This feeds the reputation system without the user ever thinking about reputation systems.
5. Getting Paid
This is where we either win or lose trust permanently. Payment must be:
Fast (same day at minimum, ideally instant on task approval)
Visible (earnings dashboard that's satisfying to look at)
Withdrawable to familiar rails (mobile money, MTN MoMo, etc. traditional banks before crypto wallets)
The Bittensor/token layer sits underneath this entirely. The user sees "₦4,500 earned" not "0.003 TAO." The conversion and wallet mechanics are abstracted. Over time, as users become more comfortable, you surface the option to hold earnings as tokens, stake, and participate more deeply but this is a later chapter.
6. Growing on the Platform (The Bittensor Layer, Made Human)
Users shouldn't encounter Bittensor as a concept until they're already earning and engaged. Then it's introduced as an answer to a natural question they'll have: "Can I earn more?"
Frame it as:
The more you contribute to Flow, the more Flow works for you.
Or every task you complete builds your reputation and your reputation earns for you, even when you're not working.
Or do great work. Get recognized. Earn more over time.
Introduce the token layer through a "Flow Score", a visible, growing number that represents their cumulative contribution to the network. Flow Score is about reputation, a running total of user's overall contribution to the platform over time, which eventually unlocks token rewards. It's not per-task pay, it's more like a loyalty/status layer that sits on top of earnings.
Functionality Priority List
Phone/email signup with SSO options
Conversational Jarvis onboarding that builds knowledge base
Skill-matched task feed with Jarvis explanations
Task acceptance, in-app work submission, completion flow
Earnings dashboard with mobile money withdrawal
Basic Jarvis messaging (handles communication on user's behalf)
File and link uploads to enrich Jarvis knowledge base
Jarvis autonomy progression (tracks trust, gradually expands what it handles)
Task decomposition visibility (show users which part of a larger task they're doing)
Ratings and reputation display
Flow Score introduction
Push notifications for task matches and payments
Token wallet and Bittensor integration surfaced to active users
"Sponsor a learner"
Multi-party task coordination (teams on WorkStream)
Advanced Jarvis: autonomous bidding on tasks, negotiation
Staking and passive income features
Key Design Principles to Hold Through the Process
Things to Watch (From the Doc)
The cold start problem is real and should inform your MVP scope. We're not trying to build all three systems at once.
WorkStream with a simplified Jarvis (smart profile + matching, not full autonomous agent) is the right first product.
Bittensor integration can follow once there's transaction volume to generate meaningful signal.
The autonomy gradient is a UX design problem as much as a technical one. We have to build the trust progression visually so users can see Jarvis "leveling up" alongside them. Make it feel earned.
The value attribution problem (splitting pay across contributors) needs to be solved before we launch multi-contributor tasks. Users will accept a lot of friction in a product, but they will not accept ambiguity about money.
The private knowledge vs. shared intelligence tension is the most important thing to communicate clearly in your privacy/trust UI. Users need to understand that Jarvis keeps their knowledge private, and that only anonymised signals contribute to the network. This should be a visible, prominent guarantee, not buried in a terms of service.
