Kitteh: tendermint in Rust

Why rust?

• C is too slow to write this much code rapidly
• C++ is dying, outside of the Boost libraries
• Go presents problems
  • CGO FFI
  • memory leaks
  • it is necessary to use Go in a non-idiomatic manner when building a blockchain, because we require determinism
    • goroutines aren't good for deterministic software
    • channels aren't
  • emxample
• Better low level library support
• Lower memory consumption for blockchain applications in production
• More rigorous compiler
• V, Nim, and Zig aren't mature enough yet

Theses

• IBC or similar proof based systems will be used for:
  • blockchain scaling
  • safety
  • communication between distinct networks
• Rust is the most suitable language for:
  • blockchains
  • static blockchain applications, like the cosmos-sdk
  • smart contracts
• Substrate/Grandpa/BEEFY lack the clear finality guarantees that tendermint brings
• In every known case, when building these systems in Go, we encounter:
  • Issues with the garbage collecgtor
  • Issues with CGO FFI
  • Memory Leaks
• Tendermint can be ported to rust for less than $1,000,000 in less than one year
  • The "surgical" practices of Informal Systems result in the dumb kind of conservatism
• Context switching between Rust and Go as currently seen throughtout Cosmos is extremely harmful to developer productivity.
  • This also seems to affect the productiivty of humans leveraging LLMs and such
• AI systems will soon do most programming tasks, but they'll require skilled humans to verify the code.

Tools

Kitteh should be primarily developed by humans using LLMs. This approach will let us move more rapidly than our counterparts, and deliver software at a lower cost than our counterparts.

To ensure the growth of an ecosystem around kitteh, we should document our techniques. Within 2 years we should anticipate the development of systems that are able to read, understand, and optimize kitteh's codebase.

• OpenAI 128k token "turbo" *
• xAI
• 01ai
• cursor editor

What shouldn't change

• ABCI
• Consensus

What needs to be removed

• 26657
  • the RPC & websocket

What needs to change

• tight coupling of consensus and p2p -- consensus code should not be in p2p and vice versa
• the current p2p stack in favor of libp2p