Epic: Lurk and Filecoin Green collaboration

[jpeg07]

Working to prototype a system in Lurk allowing a project to model its energy use and make representations to customers.

• Outline of an MVP and notes doc: https://docs.google.com/document/d/11W3y2yhct75gJmvlLbexyT1Dyn-HBs_BiROGKNMkgF4/
• zk-Emit Google slides: https://docs.google.com/presentation/d/1kCdiMhq_SHVcZn_bhLDj7RP6p7MC9NmKef_dpCyI5rU/

Relevant people: @redransil @mjohnson518 @porcuquine @emmorais @jpeg07

[jpeg07]

Comments from @porcuquine in Slack, July 28:

Maybe there is a better place for this message (let’s decide on one and create), but while on my mind: I think maybe the easiest way to deal with the ‘multiple time steps’ issue we discussed for an MVP would be if we can keep the model small enough that one time step can reasonably fit into a row (even a very large row).

The top row can contain initial inputs, and each row’s formulae can refer either to the previous or current row (while respecting public/private boundaries).

Each ‘organization’ can have a range of columns, further subdivided into a public and private part.

Alternately, if it’s easier to manage or reason about, we could swap rows and columns. Each org has a range of rows. If the model gets large, it just has a lot of rows.

[jpeg07]

Pasting the "notes" section from the MVP doc here:

Notes

• Bounds on emissions for different types of emissions models
  • Electricity, electric arc furnaces, etc -> trigger audit
• DAO needs subject matter experts who are involved in adjudicating disputes
• Auditors could become a bottleneck, particularly because they need to be experts in both GHG accounting and Lurk
• Aggregate intensity/production information for an industry
  • Or opt out of reporting a variable?
• “Framework” implements the global spreadsheet. “Application” is an individual model
• Need to come up with a specific MVP application that implements all of the MVP features above
  • Example: SP who stores data and provides both estimates and proofs as they resolve
• What specifically does the model need to be capable of, computationally
  • Big int? Negatives? Floating point?
  • Does field elements rn
  • Less than and greater than aren’t currently cheap
• Practically how to set up
  • (1) Inputs: write some sort of data scraping tool that gets the data we need from the external world and provides it, simulating individual data providers
  • (2) some application level program represents the entity (ie company or project), receives data and uses the library or tool to produce an output file and proof representing entity X (ie some data source) contribution to the next clock cycle
    • And produces the proofs
    • Could separate into two entities: one that uses the lurk system, and the other that is the lurk system itself (application and framework)
• Put two systems together
  • One google doc that describes what we are doing and what we think we’ve done
  • Second a spreadsheet model that actually does this as the example
    • Just need to model one clock cycle unambiguously (either two different regions of one sheet, or just set up a sheet that internally has all the relationships)

[jpeg07]

@redransil - my understanding is that the spreadsheet we need for the next steps here (bold, above) will come from your team. Let me know if we are on the same page here. Thanks!