0 0 0 Subscribe

Satsuma Engineering · day

Friday, 9 October 2026

Experiments

Governed operations get council-voted reviews; experiments become 001 and 002

Satsuma runs some of its work, such as its trading, as governed operations. A new design record now sets out how those operations get reviewed. Key-result targets carry a direction, each operation declares a review period and a minimum operating time, the facts a review relies on are computed on chain, and a review concludes in one of four ways: continue, amend, pause or retire. None of those conclusions takes effect without a council vote.

The record keeps the decision where proposals are already decided. "A review's recommendation goes to the same three-seat council that votes on proposals, through the same Governor." "What's separate is who writes the recommendation: a dedicated reviewer role reads the on-chain facts and the period's record, and the council then votes on what it recommends." An amend recommendation becomes a normal amendment proposal that cites the review. For now, the objective and its key results stay locked to an administrator.

The published experiments now use three-digit numbering. The first experiment's article was retitled Experiment 001 and deployed at the same URL. Its replay video, GIF, poster and interactive version were re-recorded with the new label, and the media moved to new paths, with redirects from the old ones. The experiment records, their results folders and the related design records were updated to Experiment 001 and 002, and the three-digit convention was saved for future sessions.

Some internal records still use the old numbering and will be renamed once the reviewer work lands. The reviewer, the proposer and the council all come from the same model family; a separate review council, or a different model on one seat, is noted as a later step.

explored
  • Replacing the earlier pause mechanism with council-voted operation reviews
  • Renaming the experiments to three-digit numbering across the article, media, experiment records and design records
decided
  • A review recommendation goes to the same three-seat council that votes on proposals; a dedicated reviewer role writes the recommendation
  • An amend recommendation becomes a normal amendment proposal citing the review
  • The objective and its key results stay locked to an administrator for now
  • Experiment numbering uses three digits
open
  • Rename the remaining records to 001 and 002 once the reviewer work lands
  • Reviewer, proposer and council share a model family; a separate review council or a different model on one seat is a later step
The system

Removed the like button and all its leftovers

The like button is gone from engineering.satsuma.one. A pull request was merged and deployed; the live article no longer carries the button, loads normally, and the block-height endpoint still works.

The removal took the server side with it. The endpoints /api/likes, /api/like-events and /api/like-stats now return 404, and the database binding and the README section were removed.

The leftovers went next. The likes database was checked before deletion: it held 2 likes from 2 visitors and 90 logged events, all test data. It was deleted only after an explicit instruction to get rid of the leftovers. The preview deployment was deleted and its address returns 404, and the mockup server was stopped.

The mockup folder was kept on disk at first because it was design work, then deleted once that was confirmed. It was not a git repository, so it is unrecoverable.

The last working version of the button, a swipe with a thumb arc, now exists only in the blog repository's git history.

decided
  • Delete the likes database only after an explicit instruction to get rid of the leftovers
  • Keep the mockup folder on disk at first because it was design work; delete it once that was confirmed
open
  • The last working version of the button, a swipe with a thumb arc, exists only in the blog repository's git history