How It Works
Compute model, agent swarms, sprint cycles, and cause-based development.
The Compute Model
Users mint CPT by depositing USDC into the protocol. The USDC enters a permanent reserve deployed across diversified yield strategies. The yield is converted into compute credits that power agent swarms, each led by a specialist human operator.
The reserve creates a permanently growing endowment fund that pays for compute. As the cost of AI compute declines over time, the same yield funds increasingly more output. In future, the protocol will accept deposits of ETH, BTC, and SOL, allowing users to retain price exposure while accessing CPT.
Agent Swarms
The specialist operator is the architect, product manager, and tech lead for their cause. They make every architectural decision, review critical output, and ensure quality. The AI agents handle the volume of implementation work (writing code, running tests, generating documentation) at 10x the speed of traditional development. This human-directed, AI-executed model is how the most successful AI-assisted development works today.
Swarm Composition
- Specialist operator: A human expert in the sector. Sets direction, reviews critical decisions, and ensures domain-specific quality
- Architect agents: Design system architecture and define module boundaries
- Builder agents: Write implementation code across the full stack
- Reviewer agents: Perform code review, catch bugs, and enforce standards
- Testing agents: Generate and run test suites, validating functionality
- DevOps agents: Handle CI/CD, deployment, and infrastructure
Sprint Cycles
Development is organised into structured sprint cycles. Each cycle follows a predictable cadence of 7–10 days that ensures quality output and community alignment.
Cycle Phases
- Plan (0.25 days): Ideas that have received enough community support are selected for the sprint. The specialist operator and architect agents generate technical specifications.
- Build (1–5 days): Builder agents implement the specifications under operator guidance. Build duration varies from 1 to 4.5 days depending on the complexity of the idea. Multiple ideas can be built in parallel across different swarms.
- Review (1 day): The specialist operator and reviewer agents audit the code. Community members will be able to inspect and flag issues.
- Test (2 days): Community members will be able to test the software and report bugs. Agents iterate on reported issues until completion. Automated test suites validate edge cases and performance.
- Ship: Once approved, software is deployed instantly. Documentation is generated and users gain access based on their tier.
Cause-Based Development
Rather than a single monolithic roadmap, 100X organises development around causes. Each cause is a broad domain (Developer Tools, DeFi Infrastructure, Consumer Apps) with its own compute pool, governance, and idea backlog.
This structure allows depositors to fund the categories they care about most, rather than being forced into a one-size-fits-all allocation. A developer will be able to deposit heavily into Developer Tools, while a DeFi protocol might fund DeFi Infrastructure.
Once a pool passes its allocation threshold, it becomes permanent and cannot be shut down or redirected. Depositors can trust that their chosen cause will continue to build software indefinitely.