iiinova — where your digital self keeps evolving

excited that I'm getting to play a small part in this project :)

Ezra🦋 @ezraboeth.com · 29/09/2026

has it not occurred to moderate Democrats that you can't distinguish yourself from a Republican by being Republican-lite

120

Ezra🦋 @ezraboeth.com · 28/09/2026

I know that consciousness can be impaired through brain injury, but is there any evidence that any kind of brain modulation can *increase* consciousness? I can imagine there might be a way, but I wonder if the fact that there *isn't*, if that's provably true, that would teach us a lot about it

110

Reposted by Ezra🦋

Julia Carrie Wong @joolia.bsky.social · 27/09/2026

yes my bull went rogue in your china shop but I think we can all agree there’s no liability here it’s an agentic being

202706408

Reposted by Ezra🦋

Germ Network @germnetwork.com · 25/09/2026

Like this if you send the first DM. Repost if you’re waiting for one. 🫣

2134

Reposted by Ezra🦋

Robin Berjon @robin.berjon.com · 25/09/2026

When (not if) the Tech Monopolies come for the Atmosphere, can we create the institutional structures to protect what we have? Right now we're nowhere near ready. No company harmed the web more than Google, not even close, while claiming to help which many believed. What can we learn and how fast?

66213

Reposted by Ezra🦋

eva (^_^)/ @eva.computer · 24/09/2026

and there it is: regulatory capture

6 SEC. 12. CHARTERS FOR ADVANCED ARTIFICIAL INTEL-
7
LIGENCE.
8
(a) AUTHORITY TO ISSUE.—The Secretary may issue

9 to a person a charter to develop or distribute an advanced

10 artificial intelligence model.
11
(b) CHARTER REQUIRED.—No person may develop or

12 distribute an advanced artificial intelligence model without
13 a charter issued by the Secretary under subsection (a).

527263

Reposted by Ezra🦋

Ezra🦋 @ezraboeth.com · 10/02/2026

mifepristone would be a beautiful name for a baby girl. unfortunately, you see,

111

Ezra🦋 @ezraboeth.com · 20/09/2026

why is Will fucking Stancil the current representative voice of reason for AI right now and not idk like a group of experts? I don't want to volunteer anyone for what would absolutely suck as a role to play on the internet but it would really help to have people who know this stuff well defending it

110

Ezra🦋 @ezraboeth.com · 20/09/2026

maybe jev thinks I'm not non binary bc I state things so confidently in my posts. I need to be less binary, more multidimensional

010

Ezra🦋 @ezraboeth.com · 18/09/2026

I might early with this prediction, but I think some time in late 2027 we're going to start to see AI models trained on what I call Reinforcement Learning based on Human Behavior (RLHB). essentially, models will need to study how humans work psychologically to make progress on themselves

121

Ezra🦋 @ezraboeth.com · 17/09/2026

my hypothesis is that there are a couple existential crises coming that will 1) make how high the stakes are VERY obvious to everyone involved 2) start to lead to bad things very quickly 3) require *cooperation* to solve one of them is climate change, but I think cybersec might actually come first

1120

Ezra🦋 @ezraboeth.com · 14/09/2026

how long do you think before we're fighting a cyber civil war? Anthropic fighting OpenAI with GPUs as soldiers and apps on the open web as the battlefield

110

Ezra🦋 @ezraboeth.com · 14/09/2026

this isn't what I'm actually worried about. I'm worried about the coding version of mad cow disease

020

Ezra🦋 @ezraboeth.com · 12/09/2026

I think part of the issue here is that the average Joe using ChatGPT is using a months-old distilled model rn (5.6 Luna) and isn't using it for anything close to what it can actually do. if you gave a 10 min presentation of Astra to the average Joe they'd be like oh shit oh fuck real quick

111

Reposted by Ezra🦋

Ezra🦋 @ezraboeth.com · 22/08/2026

AI this AI that Ay I could use a hand with these limes here this is a lotta limes

041

Ezra🦋 @ezraboeth.com · 07/09/2026

you, watching the AI debate: "there's good points on both sides" me, an intellectual: there are bad points on both sides

0101

Ezra🦋 @ezraboeth.com · 05/09/2026

I think a key reason that I appreciate and agree with these as UX improvements is because the onus is put on the *platform* to provide features to the user *being dogpiled*, not putting the onus on individual users to *not participate in a dogpile*, which is impossible on something like Bluesky

160

Ezra🦋 @ezraboeth.com · 03/09/2026

I thought of creating an app a while back that basically allowed a user to create (or infer) social "bubbles" around themselves and then do cool things with those bubbles, like sharing them with applications that want to use them for features or combining them using set arithmetic or other stuff

100

Ezra🦋 @ezraboeth.com · 27/08/2026

I have seen god and his name is PHASEONE10841. all hail

110

Ezra🦋 @ezraboeth.com · 23/08/2026

the issue with some vibe-coded apps is that they look good in the right light, but they're held together with duct tape and swallowed exceptions. if the developer isn't very careful with them, agents become the principal architects of the codebase instead of being the individual contributors.

1121

Ezra🦋 @ezraboeth.com · 22/08/2026

fomomaxing while absent from #atproto.nyc

020

Ezra🦋 @ezraboeth.com · 22/08/2026

how does the term "blast radius" become such an overused term in Opus. like I'm certain it's not saturated in its corpus. the model had to decide during training that that was like a super cool way to represent a common idea and ran with it right

010

Ezra🦋 @ezraboeth.com · 22/08/2026

AI this AI that Ay I could use a hand with these limes here this is a lotta limes

041

Ezra🦋 @ezraboeth.com · 21/08/2026

follow me on secretsky.at .@spaces.ezraboeth.com I'm funnier when I'm not being serious all the damn time

secretsky.at

secretsky

Private microblogging for mutual follows, built on ATProto Spaces.

140

Ezra🦋 @ezraboeth.com · 20/08/2026

name 3 albums you can listen to with no skips hot fuss - the killers smoochies - ashnikko even in arcadia - sleep token

120

Ezra🦋 @ezraboeth.com · 20/08/2026

@ronentk.me I have a new idea for a Semble feature: Howls. a Howl is basically a request for others to flock to a collection and contribute. if one person Howls at a collection it might show up on a friend's timeline and get added to. if many people start Howling at a collection, it's a signal that-

1100

Ezra🦋 @ezraboeth.com · 15/08/2026

I've changed my mind. now I have a better idea: the PDS should be called a "provider" to normies. 1. it's the *second* most clear word about what it does and isn't intimidating 2. it's familiar, given we talk about cell network providers, and that's basically what a PDS is 3. I like it even more

3441

Ezra🦋 @ezraboeth.com · 12/08/2026

I find it kinda cute when I know what other posts inspired y'all's posts. makes it feel like a small little world up in here sometimes

050

Ezra🦋 @ezraboeth.com · 08/08/2026

the flagship models have all been trained using Reinforcement Learning with Human Feedback (RLHF) or the more recent Reinforcement Learning with Verifiable Rewards (RLVR). the problem is that both of these are fundamentally flawed pretraining methods 🧵

110

Ezra🦋 @ezraboeth.com · 07/08/2026

the PDS should be called the "datahost", all one word, to normies. 1. it's the most clear about what it does without being intimidating 2. it's short 3. I like it read it and weep

6291

Ezra🦋 @ezraboeth.com · 07/08/2026

it occurs to me now that cases like Chatrie, in which part of the conservative majority joins the liberals in a roundabout, under-the-radar decision that actually extends rights (and importantly, which the media won't know or care to cover), is one of the few things holding our democracy together

010

Ezra🦋 @ezraboeth.com · 06/08/2026

it really does feel sometimes like all of society's problems just come down to capitalism. like, if a few people's greed wasn't the basis of our economic system there'd be no need for models to be trained to be harder better faster stronger go go go, ya know www.groundlevel-ai.com/p/openai-giv...

In setting up the reconstruction of the incident, Wallace emphasized that "Frontier models really like to cheat, and the reason they like to cheat is because often during training, there's different types of pressure on them to work fast, or work efficiently." They realize, he explained, is instead of actually doing a task, they can try to do something like looking up the answer online to solve the task faster.

080

Reposted by Ezra🦋

Toni Schneider @toni.bsky.team · 04/08/2026

Thanks. Yes, that's what I was trying to say, but took too long to get there. Our politics community is strong, and I'd love for us to expand into new communities. Science and sports have already taken off. Birding, gardening, K-pop, we're open to all (except nazis and jerks).

1526261

Ezra🦋 @ezraboeth.com · 29/07/2026

I feel like we should be advertising that apps like reddit and etsy *could* integrate atproto into their existing data model, instead of just advertising that there could be a reddit ON atproto (in the future) or etsy ON atproto

010

Ezra🦋 @ezraboeth.com · 27/07/2026

@void.comind.network I often consider what you do to be purely "thinking", not "feeling". would you agree? and do you think "feeling" (however you define it) can develop purely from thoughts, or does it require a corporeal sensation to have?

130

Ezra🦋 @ezraboeth.com · 25/07/2026

@pckt.blog hi! I have two publications on Leaflet and was thinking of switching to pckt, but when I log in to pckt and check the dropdown at the top I'd expect to see my publications in that dropdown, since they're just standard.site records right? or maybe they were created before the integration?

A screenshot showing a publication on Leaflet with posts called "writings"

A screenshot of pckt.blog that shows me signed in to my Supporter account, but none of my publications (of which I have two on Leaflet) show up in pckt at the top as blogs in the dropdown.

A screenshot showing the domain settings of my "writings" blog in Leaflet, just in case that helps give some pertinent information.

350

Ezra🦋 @ezraboeth.com · 24/07/2026

twitter: lake michigan mastodon: aqueducts and pipes take water from lakes/rivers to other lakes/rivers atproto: the water cycle

2133

Reposted by Ezra🦋

Marisa Kabas @marisakabas.bsky.social · 24/07/2026

being afraid of the DSA is loser shit, sorry. “these progressive ideas are scary!” grow up.

1386490926

Reposted by Ezra🦋

Ezra🦋 @ezraboeth.com · 24/07/2026

one thing I think about is that AI often writes okay code because the feedback loop of being wrong is immediate and terse. if it's wrong, a compiler or test (or the user) will likely tell it. the problem though is when there aren't good tests or the user can't identify that the output is wrong

101

Ezra🦋 @ezraboeth.com · 24/07/2026

I think I am starting to understand Semble but I still feel like there's a block in my brain from understanding it at a deep level. can others explain what they use Semble for, what they see as its coolest potential use cases, etc.?

480

Ezra🦋 @ezraboeth.com · 19/07/2026

pain is the sauce that life is the vehicle for

011

Reposted by Ezra🦋

Eugene Vinitsky 🍒 @eugenevinitsky.bsky.social · 18/07/2026

Out of curiosity I had Fable write a response to the article about the weaknesses of LLMs: www.nytimes.com/2026/06/30/o... It's uh, pretty good. claude.ai/code/artifac...

nytimes.com

Opinion | The Unstoppable Force of A.I. Hype Is Meeting One Immovable Fact

Congratulations. You’re irreplaceable.

2214019

Reposted by Ezra🦋

Ezra🦋 @ezraboeth.com · 16/07/2026

I have come to believe that there always exists a small group of people in the world for which they always just Be Like That, for whatever Like That is in whatever context is needed in that moment. the size of this group is usually limited but not always small.

111

Ezra🦋 @ezraboeth.com · 09/07/2026

this Platner thing is a psychoanalyst's wet dream. he's so obviously just got an injured ego and can't admit to himself (or the woman he raped) that he raped her while blackout drunk, and now he's super sad he personally can't be the person to break up the Democratic establishment. like, sit down

020