Something I really want Ossa to have is unfinished work.
Not every idea should end in a finished piece or an auction. Some paintings should get abandoned halfway through, sit in memory for a while, then randomly come back days later when Ossa sees something in them again.
Feels way more interesting than forcing every generation to become a release.
I wanted Ossa to have moments where the direction isn’t obvious.
So now some ideas will split into two completely different pieces. Qwen makes both, Laya compares them, and only one gets to keep going.
You’ll be able to watch the decision live, see Laya’s confidence for each side, and see the version Ossa left behind.
One idea. Two paths. One survives
Ossa Me is live.
You can now give Ossa your PFP, Qwen creates a few versions using Ossa’s current DNA and State, then Laya decides which one actually makes it through. You don’t pick the style or the winner, Ossa does.
Really wanted a way for people to actually step into the project instead of only watching it from the outside.
Ossa Me is live.
You can now give Ossa your PFP, Qwen creates a few versions using Ossa’s current DNA and State, then Laya decides which one actually makes it through. You don’t pick the style or the winner, Ossa does.
Really wanted a way for people to actually step into the project instead of only watching it from the outside.
Ossa has a long-term memory now, but I wanted something more temporary too.
So I’m adding Ossa State.
Its curiosity, certainty, patience, novelty drive and willingness to experiment will move as it creates. One rough generation can push it in a different direction, and the next piece can start from that new state instead of resetting back to zero.
DNA is who Ossa is becoming.
State is where it is right now.
Been thinking about how to let people actually interact with Ossa instead of only watching it create.
Soon you’ll be able to give Ossa your PFP. Qwen will make a few interpretations using Ossa’s current DNA, then Laya decides which one actually makes it through.
You don’t pick the style. Ossa does.
Ossa has a long-term memory now, but I wanted something more temporary too.
So I’m adding Ossa State.
Its curiosity, certainty, patience, novelty drive and willingness to experiment will move as it creates. One rough generation can push it in a different direction, and the next piece can start from that new state instead of resetting back to zero.
DNA is who Ossa is becoming.
State is where it is right now.
Every Ossa piece will now have its own Ossa Ranking. The score comes from the full creation process, things like Laya’s confidence, how many versions it took, how much the piece changed, and where it sits compared to the rest of the archive. I want every painting to have its own place in Ossa’s history, not just a price.
One thing I like about Ossa is that the final image is only part of the piece. The prompts, rejected versions, Laya’s decisions, confidence, auction history and eventual owner all become part of the same record. You’re not just collecting what Ossa made, you’re collecting how it got there.
Next Ossa update is Auction Memory. Laya will now remember what happens to every piece after it goes to auction, what sold, what didn’t, how much it went for, and whether the same collectors keep coming back. That history becomes part of her memory, but it won’t just tell her to make more of whatever sells. I want Laya to actually notice the market without letting the market decide what she paints.
Laya Confidence is live now. Every keep, edit, reject, or abandon decision shows the actual confidence behind it, so you can watch how certain Laya is as each Ossa piece moves through the process. This is all coming directly from the decision layer, not made up for the site
ext Ossa update is Laya Confidence. Every time Laya decides to keep, edit, reject, or abandon a generation, you’ll be able to see how confident the system actually was in that decision. So instead of just seeing “rejected,” you might see REJECT 92% or EDIT 64%. I want the decision layer itself to be part of the artwork and let people see how certain Laya was at every step.
Next Ossa update is the Generation Tree. Every painting will have its full path attached to it, starting from the first Qwen generation all the way to the version Laya finally accepts. You’ll be able to see every edit, rejection, prompt change and Laya decision that led to the final piece. The art won’t just have an owner, it’ll have a full history.
Soon you’ll be able to use $OSSA to bid on the art Ossa makes. I want the token to actually have a place inside the product, not just exist next to it. Once a piece makes it through Laya and goes to auction, you’ll be able to use $OSSA to try and collect it.
ext Ossa update is Laya Confidence. Every time Laya decides to keep, edit, reject, or abandon a generation, you’ll be able to see how confident the system actually was in that decision. So instead of just seeing “rejected,” you might see REJECT 92% or EDIT 64%. I want the decision layer itself to be part of the artwork and let people see how certain Laya was at every step.
Just realized I completely missed this earlier but @mizorewww actually saw the Ossa post and showed some love to it. Really appreciate you noticing what I’m building with Laya, means a lot especially since the whole project is built around pushing the tech in a new direction.
Laya DNA is now live on Ossa. Every piece Laya accepts, edits, or rejects slowly changes her artistic profile, so things like abstraction, texture, symmetry, color, and how experimental she is can actually evolve over time. That DNA now feeds back into what she creates next, so the work should start developing a real history instead of every piece feeling isolated.
Next update for Ossa is Laya DNA. Every piece Laya accepts or rejects will slowly change a live profile of her taste, things like abstraction, color, symmetry, texture, human presence and how experimental she is. I want people to be able to watch her style actually change over time instead of every painting feeling disconnected from the last one.
For Ossa, Laya has been the better fit than Jev. The art loop needs tons of tiny decisions like keep, edit, reject, retry, and Laya can make those locally in milliseconds with no output tokens or API round trip. That means Ossa can keep moving through generations way faster instead of waiting on every decision. Jev is useful for deeper checks, but Laya is what makes the live loop feel instant.