Status: Exploratory research — no decisions made
Skychain is a proposed Sky-native EVM blockchain designed for AI agents, not humans.
Core Principles
Performance First
Optimized for throughput and efficiency. Without human UI constraints, we can make different trade-offs.
State Rent
Storage costs ongoing fees. Expired state is gone — no recovery mechanisms needed.
AI-Native
No human UI trade-offs. Agents can handle complexity that humans can't.
Why AI-Native Changes Everything
Traditional blockchains make trade-offs for human usability:
- Friendly error messages
- Forgiving interfaces
- State preservation and recovery
- Human-readable transaction formats
- Wallet UX considerations
Skychain doesn't need these.
Agents can:
- Handle state rent complexity (track expiration, re-create as needed)
- Work with minimal/no UI
- Tolerate stricter operational requirements
- Optimize for machine-readable interfaces
- Manage their own state lifecycle
State rent without recovery is acceptable because agents don't need human safety nets. This unlocks aggressive state pruning that traditional chains can't do.
State Rent Design Space
Why State Rent?
State bloat is the existential problem for blockchains. Every transaction adds state that must be stored forever. This doesn't scale.
State rent means: storage costs ongoing fees. If you stop paying, your state expires.
Design Options
Rent Models:
- Per-byte periodic fees
- Prepaid deposit that depletes over time
- Auction-based storage pricing
- Tiered pricing (hot vs cold state)
Expiration Behavior:
- Hard delete (state gone forever)
- Archive to cold storage (retrievable but slow)
- Grace period before deletion
- Tombstone markers
For AI-native: Hard delete is acceptable. Agents can track their state and re-create if needed. No need for recovery complexity.
Research Questions
- What's the optimal rent price curve?
- How do agents coordinate around state lifecycle?
- What's the minimum viable state rent implementation?
- How does rent interact with DeFi primitives (vaults, positions)?
Base Chain Options
Skychain needs a starting point. Options to research:
Monad
- Performance: Parallel EVM, 10k+ TPS claimed
- Architecture: Optimistic parallelism, pipelined execution
- State rent: Would need custom modifications
- Questions: How does parallel execution interact with state rent? Can we fork it?
Cosmos SDK + Ethermint
- Performance: Proven, modular architecture
- Architecture: Tendermint consensus, IBC for interop
- State rent: More native support possible in SDK
- Questions: Is Ethermint EVM compatibility sufficient? IBC bridging to Ethereum?
Reth Fork
- Performance: Rust implementation, active optimization
- Architecture: Clean codebase, modular design
- State rent: Would need deep modifications
- Questions: How much work to add state rent? Performance vs Geth?
Sei v2
- Performance: Parallel + optimistic execution
- Architecture: Twin-turbo consensus
- State rent: Custom work needed
- Questions: How does their parallelism compare to Monad?
Custom L2 (OP Stack / Arbitrum)
- Performance: Inherits L1 security, customizable execution
- Architecture: Proven L2 patterns
- State rent: Easier to customize execution layer
- Questions: Can state rent work on an L2? Settlement implications?
Performance Considerations
Without human UI overhead, what's the performance ceiling?
Things we don't need:
- Human-friendly RPC responses
- Transaction simulation for wallets
- Gas estimation UX
- Mempool transparency for frontends
- Block explorer optimization
Things we optimize for:
- Raw throughput
- Latency for agent transactions
- Efficient state access patterns
- Machine-readable everything
- Batch operations
Research Questions:
- What TPS is achievable with AI-native assumptions?
- What's the latency floor for agent operations?
- How do agent transaction patterns differ from human patterns?
Synome Integration
How does Skychain fit into the broader Sky ecosystem?
Connection to Synome
- Could Skychain be the native substrate for Synome operations?
- State rent parameters governed through Synome?
- Sentinel operations across Skychain?
Relationship to Ethereum Mainnet
- Sky currently lives on Ethereum mainnet
- How would Skychain bridge to mainnet?
- Asset movement between chains?
- Which operations stay on mainnet vs move to Skychain?
Cross-Chain Coordination
- Could Sentinels operate across both chains?
- How do rate limits work cross-chain?
- Synome as coordination layer above both?
Research Agenda
Phase 1: Landscape
- Survey existing state rent proposals (EIP-4444, etc.)
- Analyze base chain options in depth
- Understand agent transaction patterns
Phase 2: Design
- Propose minimal viable state rent model
- Identify base chain fork candidate
- Design Synome integration points
Phase 3: Prototype
- Build proof-of-concept
- Benchmark performance
- Test agent workloads
Discussion Topics
State Rent:
- "What's the simplest state rent that works for agents?"
- "How aggressive can expiration be before it breaks DeFi primitives?"
- "Should rent be flat or dynamic based on demand?"
Base Chain:
- "Monad vs Reth fork — what are the real trade-offs?"
- "Is L2 the right approach or do we need L1?"
- "How important is IBC/interoperability?"
AI-Native:
- "What human UX assumptions are baked into current EVM?"
- "How do agent transaction patterns differ from humans?"
- "What performance is possible without wallet UX constraints?"
Integration:
- "How would Sky assets move between mainnet and Skychain?"
- "Could the Synome span multiple chains?"
- "What stays on mainnet vs moves to Skychain?"
This Is Research
No decisions have been made. We're exploring the design space.
What we're looking for:
- Deep analysis of options
- Trade-off comparisons
- Hard questions that reveal constraints
- Novel ideas we haven't considered
What we're not looking for:
- Advocacy for specific solutions
- Surface-level comparisons
- Hype or speculation
Explore, question, analyze. That's the work.