(which prevents Coinbase from appropriately forming a rebuttal to the Wells Notice since they don't know which instruments they're defending) pretty disappointing to find out that the securities regulatory body of the U.S. has a fundamental conflict of interest wrt general public's ability to innovate in a market.
I often think about this wrt to subsistence homesteads. if there was a fungible hypercert reward for every group honestly tending the land & building a living structure, they're probably eligible for hypercerts for extending that to off-grid energy + climate control + on-premise compute. effectively the entire thing.
100%, imho there will be a supercritical architecture that strips out the frills and hones in on a forum/multisig/CLI with version control and maximal dogfooding. people should be searching for the "EEAAO" DAO well before any token or certificate of impact is needed.
there should be a name for the perceived time constriction of quickly developing trends and OSS techniques that elicits a feeling of imminent supercriticality.
what's the sanity check on using hypercerts to curate a creditworthiness registry? I'm one of those weirdos that believes we can figure out onchain banking without trust assumption oracles.
up to 64k token input potentially: https://arxiv.org/abs/2303.09752 can't wait to see a pytorch implementation in a week
what's the deal with computers? you put in some words, it puts out some words. sometimes the words always start with "I'm sry, I can't give you those words" *seinfeld bass riff*
I'm more partial to softer cookies, then again I bake muffins and cake on their own merits as well.
ngl I like the idea that some ecosystems want coherent validiums and are willing to design a mechanism to settle those in a gas-heavy environment. then there are ecosystems that will let the validiums sprout where they may.
don't forget the windfall of luddite commenters attempting to detect/dunk on the content and providing RLHF instead. AI - "oh no, you got me. duly noted" (can GPT-n detect specialists?)
all social media apps should be rethinking how to optimize request for comments and CI/CD past github's social network. preferably with infra like radicle. really a missed opportunity for twitter (opensource features ex post not enough). if your social media isn't a personal & global roadmap with pool of testers...
I think there's more detail in parts 2/3 about prover latency and various circuit schemas.
I found this: https://youtu.be/tg6lKPdR_e4 to be pretty informative
contests for subdivisions. crazy thought, but the prompt is a lie; instead people want a natural language CLI, or a sequence of prompts, that abstracts away the daisy-chain of however many models/services.
can @chatgpt provide random prompt challenges every day? best answer could get ranked and added to an implementation of alpaca: https://github.com/tatsu-lab/stanford_alpaca question still remains how to automate transcription longform response images, but for labelers it would be just as readable.
wrt SVB collapse recap on Moment of Zen. kind of weird that dunbar-optimal groupchats/multisigs aren't their own class of bank with their own class of creditworthiness. a lot of month-to-month costs could be covered between a group of industrious friends with shared onchain responsibility.
not exactly, this explains the "cost ceiling" (https://youtu.be/acbBeGcevok?t=531) practically, there's a toll as a discovery mechanism for defending an otherwise public good with significant building cost.
I was considering this threat with respect to the possible scope of the "GPT4 using Taskrabbit to bypass captcha" but applied to email. hashcash is one solution, but imho the higher-quality work could be baked into the browser.
http://www.certificatestemplatesfree.com/wp-content/uploads/2018/01/certificate-error-google-a-google-cert-rgUnaz.jpg right now BAT is just Brave ad spend, converted and distributed to Brave users. maybe a hot take, but public social networks like farcaster will need to be a toll good as far as malicious freeriders.
🌶️ take: Brave Browser's tokenomics should have immediately gone to certification. especially wrt infosec. ask the question "how much async correspondence does it take to procure or steal X?", then ask it again with the preface "given this much staked reputability..." if opensource does X, PTAs do X + Y