Fractile – Break Work, Build Fast

Inspiration

In today’s fast-paced digital economy, startups and growing businesses often struggle to get work done due to limited budgets and resources. At the same time, thousands of skilled individuals seek short-term, low-commitment opportunities to earn income flexibly.

We saw an opportunity to bridge this gap. What if complex work could be broken down into small, independent tasks that anyone could pick up, complete quickly, and get paid for—without needing to commit to a full-time project?

That’s how Fractile was born. A platform where:

Businesses submit a big task and total budget.
AI breaks it down into granular, independent subtasks.
Freelancers pick up micro-tasks matching their niche skills, complete them, and get paid—fast.

What It Does

Fractile is a decentralized micro-task marketplace powered by intelligent task decomposition. Key features include:

One-click Task Decomposition using AI (e.g. “Build an eCommerce website” → frontend, backend, UI design, payment gateway integration, etc.)
Open Microtask Pool where freelancers browse available subtasks and pick what suits them.
Milestone-based Payouts based on startup feedback, allowing startups to control quality and freelancers to earn instantly.
Skill-based Filtering to ensure relevant contributors see tasks in their domain.

How We Built It

We leveraged bolt's capabilities to build the platform from scratch, developing the code base through multiple feedback iterations and using enhanced prompts.

Frontend: TypeScript, React.
Backend: Supabase, TypeScript, React.
AI Agent: OpenAI API to interpret large tasks and break them into independent subtasks.

What We Learned

Task decomposition is an art. It's not just about breaking things down—it's about ensuring independence, clarity, and granularity. We explored hierarchical task planning and prompt chaining with LLMs to fine-tune the decomposition.
Coordination without commitment is possible. With good architecture and trust-based feedback loops, distributed work can scale.
The true potential of AI to leverage in various software development use cases.

Challenges We Faced

Designing truly independent subtasks without overlapping dependencies was tough. We iterated on multiple AI prompt structures and used feedback loops for better granularity.
Handling quality assurance in a crowdsourced environment—how do you ensure output matches expectations? We introduced a two-tier validation system: auto-evaluation + startup feedback.
Balancing budget allocation across subtasks dynamically—especially when scope expands—required a flexible budgeting engine.