2024-08-21: Government, Little AIs, Fridges
(If you work at FF/FFDW 🄽 links will send you to a local, editable page in the Filecoin Foundation Notion. If you are in the Filecoin ecosystem, you may be able to join the #monologue-danny Slack channel where I answer questions, take meeting bookings, talk to myself and the other voices in my head. In the glorious decentralized ocap-enabled future, such data-hoardings will be a thing of the past, but we live for now in a fallen ACL world.)
Working on:
The Governance Game
(The actual Governance Game is on github). Right up there on the list of things for me to catch up on was the Filecoin governance reforms . These got snagged (as I suspected they would, all hail my amazing prophetic skills) on building and securing the tooling for them, plus the @nucleation, but are now beginning to show results. There’s the original discussion from last year, and the FIP implementing it all here (FIPs are “filecoin improvement proposals”). I’m not sure of the new timeline, but my guess is that it will be aiming for an unveiling in Bangkok in November at Fil Bangkok 🄽.
This is a tricky transition from the Auld Ways (which was “soft consensus” for most technical changes — ie almost everybody needed to agree).
My main thinking at the moment is that there are a whole bunch of ideas kicking around the ecosystem that a) had no real way of getting approved, because they would have to gain absolute approval to pass under soft consensus, but also b) it was unclear /how/ popular they were, and with whom. It was just hard to get validation for a proposal — beyond extremely technical proposals (and as @Adrian Lanzafame has noted, even they don’t necessarily get the full oversight from SPs). I’m not sure there are many Filecoin improvements that need a FIP, but there have been plenty of people who wanted to try at a FIP, but were unsure how to getting it passed. It led to a lot of thrash.
But bootstrapping a whole new system! And securing it first!
I think — as with a lot of areas of Filecoin where many variables are flying around — I would love to be able to just sort of model the damn thing in an interactive tool of some kind.
Tiny STFIL Update
The case has moved from the police to the prosecutor. In the background, there’s been some positive noises from close to the prosecutor’s about the court’s attitude to STFIL customers; I’m not sure how much credence to give that, but I’m glad for those affected that the FIL was not liquidated before it left the police’s control.
Hey you! Take my 30 second survey about what I should be doing in this letter! I LOVE YOU xoxox
Thinking about:
Local open LLMs. I keep an eye on how well these work. The llama-3.5 and gemma-2 models are small enough to run on my laptop, and give answers that seem as good as GPT a year or so ago.
My use-case for them has always been as an intermediary between me and the waves of information that come at me. I don’t want to do that with a cloud LLM provider for privacy reasons: but I’m finding this simple prompt to rate incoming posts for me is working just as well with these new, locally-run models. I’m going to keep fooling around with this, because it would help my sanity, let alone my efficiency.
Link du jour:
How to deal with fear of new technology — refrigerated food edition.