iiinova — where your digital self keeps evolving

as ive been indoctrinating @oyster.cafe into this he started calling it stateless appviews which isnt far off i dont think. the way ive often described it is that appviews should be (potentially filtered) caches of PDS data i like the term cache here personally

150

Nelind @nel.pet · 3h

the lesser of two evils absolutely but i still think we can do better on that specific front. not with the protocol primitives themselves but with community run supporting services that can help alleviate the fundamental fact that verifiable time does not exist in atproto unless you were there

130

Nelind @nel.pet · 4h

and is one of the things i see people do a decent bit when coming from the mindset of a traditional backend as for apps a resync button. theoretically shouldnt be needed ever as long as the indexer does its job correctly. probably nice to have as an "advanced" option though. nothings perfect.

130

Nelind @nel.pet · 4h

apps not picking up records unless youve made a profile record is pretty much expected behaviour. see atproto.com/guides/lexic.... specifically the part right at the bottom about signalling records. doing this with an appview side "account" is absolutely incorrect by the atproto model though yea

atproto.com

Lexicon Style Guide - AT Protocol

A style guide for creating new atproto Lexicons

140

Nelind @nel.pet · 4h

in general anything that assumes that appviews have any kind of authority over the data itself and not just what subsection of it it will display. which is a very easy mistake to make if you think in terms of a traditional backend

040

Nelind @nel.pet · 4h

invariants. stuff like bluesky postgates. when if i backfill a post that had a postgate for a bit and then it was removed. i have no reliable way of knowing which replies to exclude (this particular one is a bigger hard thing ive written about a few times before)

130

Nelind @nel.pet · 4h

theres a bunch of subtle things people might do on accident here that undermines this if they arent careful and design too much with traditional backends in mind. stuff like assuming peoples data will always be around just because that apps client doesnt allow deleting it. various cross record

130

Nelind @nel.pet · 5h

i think my biggest opinionated atproto take is "if you cant nuke your DB and rebuild 90-99% of your app state from just backfill then youre doing it wrong". all in on PDS being the only real authoritites on state in the network

4344

Nelind @nel.pet · 7h

genius thats what

010

Nelind @nel.pet · 7h

ig i just also find the arguments for putting it under the at scheme kinda ... weak

140

Nelind @nel.pet · 7h

im still not sure how i feel about atproto spaces using at://<did>/space/... as the URI instead of the ol' ats://<did>/... more or less for the same reasons that bryan point out when he argues for ats too. it fits the precedent of URI schemes better. and using at muddles stuff a bit imo

170

Reposted by Nelind

ari 💫 @arimelody.space · 17h

i'll be so for real i cannot take the term "shrinkflation" seriously i may be cooked

1292

Nelind @nel.pet · 19h

its not even lunch and cmake is already making me wanna go home ...

040

Nelind @nel.pet · 29/09/2026

"what was most answered as a genshin hear me out?" and number two is slimes ... am i fucked up that i dont really think thats a hear me out? am i the weird one?

020

Nelind @nel.pet · 29/09/2026

so tired ... my work commute usually takes like 30-40 mins ... took 2 fucking hours today !! i was sat in a single highway exit for like 50 mins ...

1110

Nelind @nel.pet · 28/09/2026

😭 thats like the whole point though ... balancing everyones interests in an organisation together. bluesky shouldnt be left out of that ... thats like saying google shouldnt contribute to the W3C because they own the most popular browser ...

010

Nelind @nel.pet · 28/09/2026

looking forward to this going further! good set of initial members/board members too.

160

Nelind @nel.pet · 28/09/2026

regardless atproto has the general "verified data transfer up to indexers, trusted data transfer down to clients" thing and this fits fine into that model probably

020

Nelind @nel.pet · 28/09/2026

i did end up seeing that in the "post-alpha" post after digging around on the forum a bit and i thiiink it would honestly cover most of my imagined use cases. either way i think its a fine tradeoff. rebroadcast isnt quite the same necessity for spaces as it is for the public stuff.

110

Nelind @nel.pet · 28/09/2026

the duality of women. sent them both the same image about biting them

1120

Nelind @nel.pet · 28/09/2026

okay yea no permissioned repos dont have verified rebroadcast at all. by design. as to avoid a leaked commit giving too much information about the contents of the repo ... still a bit sad tho :<

130

Nelind @nel.pet · 28/09/2026

ig permissioned repos just ... dont have verified rebroadcast at all?

120

Nelind @nel.pet · 28/09/2026

oh i skipped over this when reading the proposal originally Because the digest is bound by a symmetric MAC keyed from the public ikm, anyone holding the commit can compute a valid mac for any hash, so a rebroadcast commit cannot prove what the user wrote

120

Nelind @nel.pet · 28/09/2026

mainly cause like. we cant have secondary services that offer limited/scoped but still verified access to a permissioned repo

150

Nelind @nel.pet · 28/09/2026

i just realised ... the set hash approach of atproto spaces permissioned repos means we cant really make inclusion proofs can we? thats ... a bit sad actually

180

Nelind @nel.pet · 28/09/2026

to disable old style cred auth login. either way if it *is* tranquil basically holding peoples accounts hostage thats bad we need to fix that ...

120

Nelind @nel.pet · 28/09/2026

iiii would hope its not our fault ... ive logged into my personal tranquil instance with goat before and i dont know of anything we've done tranquil side that would block it. but could be something with how your account is configured? or how margin.cafe is configured? tranquil does have some options

120

Nelind @nel.pet · 27/09/2026

another ergonomic issue is that the compiler seems to have a tough time with static dispatch here: let test: Test<_, true> = Test::new(1); is not enough for the compiler to know which implementation of new to use (theres one for both true and false). but let test = Test::<_, true>::new(1); is

010

Nelind @nel.pet · 27/09/2026

though tbf this method of doing it prooobably doesnt look amazing in the resulting rustdoc but 🤷‍♀️ kinda still wanna play around with it a bit. its an interesting approach i think

110

Nelind @nel.pet · 27/09/2026

yea. still i think this would be juuust enough to cover the motivating example i had in my head of a version of Rc and Arc that did the choice of atomic vs not-atomic as a const generic such that you could write code that works with either without having to do some per-project wrapper or something

100

Nelind @nel.pet · 27/09/2026

oh that is beautifully cursed using associated types like that. my gut tells me there are some ergonomic things with that approach that would be nicer if it was a proper language feature but like. this is a cool idea

110

Nelind @nel.pet · 27/09/2026

but yea its hard to really properly explain ig

030

Nelind @nel.pet · 27/09/2026

i remember way back when in the really early days of tangled @oppi.li asking me about "how you delete blobs". which as a question doesn't really make sense when you know how blob references interact with blob lifecycles. but i think some docs kinda lead people astray in understanding

130

Nelind @nel.pet · 27/09/2026

yea i think so too. this is kinda a thing ive been talking about even before spaces actually. that blobs kinda dont exist in their own right? theyre in some ways effectively part of the records that reference them. and talking about them as separate entities seems to be making it hard to understand

140

Nelind @nel.pet · 27/09/2026

are we defining blobs in a way thats making it hard for people to truly understand what they are and thats causing this confusion?

130

Nelind @nel.pet · 27/09/2026

i think its quite interesting the amount of people ive seen ask about blobs in relation to spaces. especially how its often in a "how can they even be permissioned at all" kinda way? i feel like permissioned access to blobs is just a natural thing together with permissioned access to records

150

Nelind @nel.pet · 27/09/2026

hmm skimming through lists of nightly rust features i cant find anything for being able to change the definition of a type depending on a const generic param? does anyone know if thats being talked about by anyone? :3 could be emulated with like. dynamic dispatch and stuffs. but would be nice still

140

Nelind @nel.pet · 27/09/2026

iiiim blind. this versions better

result of the test in the format made for sharing.

beyond the kindness result the percentages read
kindness: 79%
justice: 66%
patience: 65%
perseverance: 61%
bravery: 55%
integrity: 48%
determination: 28%

020

Nelind @nel.pet · 27/09/2026

some of these questions are a bit "so you have low self esteem dont you?" but uh yea that feels about right tbh :3

a screenshot from an undertale themed test about which of the seven souls in undertale you are most like. the result is green: kindness. the flavour text for it reads "no matter how many times i got hurt, i never let that make me cruel."

260

Nelind @nel.pet · 26/09/2026

mutable reference to a value of type Nel thats valid for a lifetime 'e :3

020

Nelind @nel.pet · 26/09/2026

i love monster hunter x vampire yuri

a screenshot of a webtoon comment reading "what a girl failure either stab her or kiss her dawgg 😭"

0120

Nelind @nel.pet · 26/09/2026

scared?? i wouldve expected a different reaction to that from you

110

Nelind @nel.pet · 26/09/2026

honestly that might be the best solution ...

030

Nelind @nel.pet · 26/09/2026

fuck you. i really cant think of any satisfactory solution that doesnt at the very least required min_specialization :< or well theres one. and it emulates the Fn trait hierarchy ... idk if i really like that solution thoooo

250

Nelind @nel.pet · 26/09/2026

a train down to the vaults? northland bank really is just gringotts: genshin edition huh

130