Notes

• FVM vs CoD:
  • FVM will be more expensive since there is a higher degree of “trust” and assurances of correctness
  • FVM is not suited for large compute jobs
  • CoD is targeting batch-like jobs (e.g. spark)
• data transfer?
  • They are currently focus on public data so they are not concerned about issues with data transfer
  • There still needs to be some thinking about the retrieval arrow.
  • Providers will get data from the IPFS gateway from the start
• Smaller scope to test:
  • Open data
  • Run on data on IPFS
  • We want to get to a model for reliable execution of compute jobs
  • They are working on a an incentive protocol already: https://www.notion.so/pl-strflt/Simulator-142fedef1e8e40848d2a6523d1e2a427
  • They also need help with the auction model
  • How to avoid providers to collude and share results?
• Current work on a simulator:
  • https://www.notion.so/pl-strflt/Simulator-142fedef1e8e40848d2a6523d1e2a427

Action items

• [ ] Write down the main problems we need to solve around incentives @Luke Marsden
  • What milestones are landing in the next 3 months
  • What econ question needs to be answered
• [ ] Review main problems and write preliminary project proposal @Anonymous

Recording

Untitled