iiinova — where your digital self keeps evolving

Reposted by Johann

Jaz @jaz.sh · 27/09/2026

Okay so I've built a new Atlas, this time it lets you explore the conversations on Bluesky over the past 7 days graphically. It should update every ~6 hours and help you navigate the information environment that exists on here. I've learned a LOT about what people do here today... atlas.jazco.dev

atlas.jazco.dev

Atlas

A living map of what Bluesky is talking about: the past week's conversations, grouped into topics and regions and rebuilt every six hours.

1171411420

Johann @johann.tezoscommons.org · 26/09/2026

The problem with the iced coffee is that if you can afford it they can't low ball you.

030

Johann @johann.tezoscommons.org · 25/09/2026

And then I tried to make a similar video and it's mind blowing.

000

Johann @johann.tezoscommons.org · 25/09/2026

Yes, but every now and then something happens that is unexpected even within the "AI is getting better at everything" expectation that kicks me out of this. The Claude song video recently was one of these things.

211

Johann @johann.tezoscommons.org · 24/09/2026

might make this into a full series going through the history and proposals / drama of tezos

120

Johann @johann.tezoscommons.org · 24/09/2026

Inspired by the Claude Song i made a little Song about the origins of #tezos

142

Johann @johann.tezoscommons.org · 24/09/2026

Interview for senior dev position in 2027: Pls explain every reference in this video.

000

Johann @johann.tezoscommons.org · 21/09/2026

Oh god I just had an idea

100

Reposted by Johann

Chris Paxton @cpaxton.bsky.social · 20/09/2026

Anthropic is working on a wet lab

516514

Reposted by Johann

mr. TIM @timkellogg.me · 15/09/2026

Jev: Fable-level model that doesn’t charge for output tokens because they’re too cheap to meter it’s not general though, it only makes decisions, doesn’t generate text, but input tokens are measured by the billion ($42/btok) typesafe.ai/blog/introdu...

Scatter plot titled "Average of 4 workflows: accuracy vs cost" comparing AI models from TypeSafe, OpenAI, Anthropic, and Fireworks across accuracy (y-axis, 40% to 80%) and cost per workflow in USD on a logarithmic scale (x-axis, $0.0001 to $1).
Data points are split into two categories: workflows (diamonds) and single prompts (circles). A frontier line highlights the most efficient workflow models—where no point is both cheaper and more accurate—connecting Jev (TypeSafe) at $0.0002 and 68% accuracy, luna (OpenAI) at $0.002 and 67% accuracy, terra (OpenAI) at $0.04 and 68% accuracy, and sol (OpenAI) at the top accuracy of 74% for $0.08. Single prompts (circles) and Anthropic models (opus 5, sonnet 5, haiku 4.5) sit below the frontier line, indicating higher cost for equivalent or lower accuracy.

2830235

Johann @johann.tezoscommons.org · 12/09/2026

We should have feeds where if you open the new post modal from there, it only posts to that feed and nowhere else.

010

Johann @johann.tezoscommons.org · 11/09/2026

and regular people are stuck at: what is log in with bsky - why? Is that a scam? or they see the link and think oh thats not bsky, and flow for users who dont have a bsky account is shit.

101

Johann @johann.tezoscommons.org · 11/09/2026

What would that look like tho, one example? I have been trying to get people to see/use this with a custom bsky frontend for our particular community, and there we can have lots of features that are great for us and regular bsky users wouldn't care, and technical people get it

300

Johann @johann.tezoscommons.org · 11/09/2026

Speaking form personal experience it is just really hard to convince people to even try something that is not the default.

100

Reposted by Johann

cee @cee.wtf · 10/09/2026

total atproto victory

1216511

Johann @johann.tezoscommons.org · 10/09/2026

nobody played so might as well give the answer: its like 5 to 10 hours depending on what numbers u use.

000

Johann @johann.tezoscommons.org · 10/09/2026

AI haters wont let you know this simple trick: "check sources on that"

000

Johann @johann.tezoscommons.org · 10/09/2026

for programming its a different use case because small errors are instantly caught by the compiler/test anyway.

000

Johann @johann.tezoscommons.org · 10/09/2026

and for each word give the meaning as it is in the sentence and pronunciation and english, deepseek sometimes puts the pronunciation into the english field and stuff like that, its a pain, sonnet never fucks up like that. (and this is a core feature of the app....)

100

Johann @johann.tezoscommons.org · 10/09/2026

deepseek, even pro, often gets things slightly wrong and its a pain if u rely for it on something where you need the results to be super consistent, for example i have chinese language learning app (good fit lol) and it needs to break a sentence down into individual words...

100

Johann @johann.tezoscommons.org · 09/09/2026

( I have such a image gen batch pending so i looked up a rough estimate, but no looking stuff up before guessing! )

010

Johann @johann.tezoscommons.org · 09/09/2026

Can you guess how much electricity it takes to generate 15k images on google cloud with a recent image model? reply with answers in x hours of this 2000 W space heater. Closest answer gets a like.

120

Reposted by Johann

cee @cee.wtf · 09/09/2026

guys making a fun game is a lot of hard work even when the computer can do literally anything you ask

61738

Reposted by Johann

ryangtanaka @ryangtanaka.bsky.social · 07/09/2026

Hey folks - happy to announce that teia.cafe has been accepted into Altalab's Fall 2026 accelerator program. I think it's a great fit for where our company is right now and also a great opportunity to get @teia.bsky.social in front of some people. www.crunchbase.com/organization...

241

Johann @johann.tezoscommons.org · 06/09/2026

But to get there it's probably easier to think about building it from scratch rather than migrating existing communities tbh.

100

Johann @johann.tezoscommons.org · 06/09/2026

I have been thinking about "fixing" the art scene today or rather what a healthy art scene would look like post gen ai. Civit is a lot closer to the future of the art industry than deviantart and old stuff, just because Civit acknowledges reality. And token gates models etc make sense....

100

Johann @johann.tezoscommons.org · 06/09/2026

I tried the higher one once and got bad side effects

250

Johann @johann.tezoscommons.org · 06/09/2026

Do you just stay on the lowest dose? Because I am doing the same for semaglutiee, but the regular schedule is increase every month and I feel like a weirdo.

160

Johann @johann.tezoscommons.org · 06/09/2026

What is your impossible goal?

100

Reposted by Johann

max power 🌌 @sneptech.bsky.social · 05/09/2026

as someone who went from UE5 to Godot earlier this year on a project, it's stupidly OP *right now* sneptech.blog/posts/2026-0...

031

Johann @johann.tezoscommons.org · 05/09/2026

Open source things like Godot are gonna be extremely OP very soon, because they are a good start for an llm to rewrite to be what u want. same with blender.

130

Johann @johann.tezoscommons.org · 05/09/2026

can u guess what model I am using for this language learning app?

010

Reposted by Johann

Brookie @brookie.blog · 04/09/2026

your bluesky account is more than just bluesky ! 😇

1395

Johann @johann.tezoscommons.org · 04/09/2026

Oh what you see is coded already just not on the main branch

100

Johann @johann.tezoscommons.org · 04/09/2026

oh and it works nicely on mobile as well, and can be added as PWA app on your home screen

040

Johann @johann.tezoscommons.org · 04/09/2026

and of course.....each group chat becomes a voice chat room and other live audio features.

162

Johann @johann.tezoscommons.org · 04/09/2026

power users can have two feeds at once, i think the regular bsky UI wastes a lot of real estate 😅

152

Johann @johann.tezoscommons.org · 04/09/2026

more discord like, for every dm / group chat we also show realted posts, could be cool if we do group chats for events like #objkt4objkt, then we can highlight those posts in the same view with the chat messages

151

Johann @johann.tezoscommons.org · 04/09/2026

We have our tezos-enabled bluesky frontend at ovoid.at, but as i add more features i have trouble fitting them into the current UI, so this is what ovoid *could* look like:

274

Reposted by Johann

alex benzer @alexbenzer.com · 04/09/2026

thread numbering was hit feature in beta, so we made it generally available. i'm personally enjoying it quite a lot — helps me decide which posts to tap in feeds. thanks to all who gave feedback during the beta. it really helped our team refine this before today's general release!

2124127

Johann @johann.tezoscommons.org · 03/09/2026

Reminder that you can get cool bsky badges on tzbsky.com

020

Johann @johann.tezoscommons.org · 03/09/2026

no, its so i can see things i like. i know the concept of liking things is confusing for haters.

000

Reposted by Johann

jeffery --dangerously-skip-permissions @jefferyharrell.bsky.social · 19/08/2026

I dislike digressing into politics, but I'll say this: The people I've encountered online who are most opposed to AI all have one thing in common and it's not their politics. They are all of them so far without exception catastrophically uninformed and really happy to be that way.

41098

Johann @johann.tezoscommons.org · 19/08/2026

This is fairly obvious when you know how training works, but still, there goes all the attribution nonese. www.csail.mit.edu/news/when-ai...

csail.mit.edu

When AI art has no author: MIT study finds generated images often can't be traced to any training data | MIT CSAIL

000

Johann @johann.tezoscommons.org · 19/08/2026

I think the bad design of hen was a feature because people would not spend time on the site other than buying and would go to socials instead for discovery. And during the hype time that worked well.

010