๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐ฐ๐ท ์ฐ์ฃผ(Would you) ๋ผ์ดํฌ $SPACE for 2026?
2026๋
๋๋ง์ TGE๋ฅผ ์๋๊ณ ์๋ Spacecoin์ $SPACE , ์ฐ๋ง์ฐจ์ ํฉ๋ฅํ๊ธฐ ์ํ ์ด๋ฒคํธ๋ฅผ ๊ฐ์ตํฉ๋๋ค!
๐ Spacecoin์?
๋ง์ค๋ฆฝ ๋ฐ ํ์ค์ ํต์ ๋คํธ์ํฌ๋ฅผ ๊ตฌ์ถํ๊ธฐ ์ํด ์ค์ ์์ฑ์ ์์ ๋ฐ์ดํฐ๋ฅผ ์ ์กํ๋ ์ญ๋๊ธ ์ค์ผ์ผ์ DePIN ํ๋ก์ ํธ. ์ด๋ฏธ 4๊ฐ์ ์์ฑ์ด ์ฐ์ฃผ๋ฅผ ๋๊ณ ์์ต๋๋ค! ํ์ฌ TGE๋ฅผ ์๋๊ณ ์์ผ๋ฉฐ, ๋ง์ ๋ฐํ๋ฅผ ์๋๊ณ ์๋ ์ํฉ = ํ๊ตญ์ @spacecoin ์ธ์ฌ๋๋ฆฌ๊ธฐ ๋ฑ ์ข์ ์ํฉ.
โญ ์ฐ์ฃผ(Would you) ๋ผ์ดํฌ $SPACE for 2026?
์ด 50๋ง๊ฐ์ $SPACE , 1์ฒ๊ฐ์ $CTC ๋ฅผ ๊ฑธ๊ณ ์งํํ๋ ๋จ๊ณ๋ณ ์ด๋ฒคํธ๋ก 1์์ค ๋ฐํ๋๋ Spacecoin ์์์ ํ์ฉํ ์ฝํ
์ธ ์ ์์. ๊ทธ๋ฆฌ๊ณ $SPACE ๋ฅผ ๋ด์ Credit Wallet ์จ๋ณด๋ฉ ๋ฉค๋ฒ๋ค์๊ฒ ๋ณด์์ ๋ถ๋ฐฐํ ์์ ์
๋๋ค!
๐จโ๐ป ์ฒซ๋ฒ์งธ ์ด๋ฒคํธ: Get Ready for ์จ๋ณด๋ฉ
ํ์ฌ 100๋ง๊ฐ์ $SPACE ์์ด๋๋ ์บ ํ์ธ์ด ์งํ์ค์ธ ๊ณณ์ด๊ธฐ๋ ํ Credit Wallet์ ์ค์น ๋ฐ ์ธ์ฆํ๋ฉด ๋. ๋น์ฒจ์์๊ฒ $SPACE ๋ฅผ ์ง๊ธํ๋ฉฐ, ์จ๋ณด๋ฉ ์งํํ ์ ์ฐฉ์ ์ฒ๋ช
์๊ฒ๋ 100% ํ์ ์ผ๋ก 1 $CTC ๋ฅผ ์์ด๋๋. ๋ฉ์ธ๋ท ํ์๋ ํธํ๊ฒ ํ๋ํ ์ ์๋๋ก ์ง์ํฉ๋๋ค.
: ํด๋น ํธ์ํฐ LIKE RTํ๊ธฐ
: CREDIT WALLET ์ง๊ฐ ์ค์นํ๊ธฐ
: ๊ตฌ๊ธํผ์ ํตํด ์ค์น ์ธ์ฆํ๊ธฐ
*๊ด๋ จ ๋งํฌ๋ ์๋ ๊ฒ์๊ธ์ ๋ต๊ธ๋ก ์์ฑํ๊ฒ ์ต๋๋ค!
๋!
์ฒซ ์ด๋ฒคํธ์ด๋๋งํผ ๊ฐ๋จํ๊ฒ ์์ํฉ๋๋ค. ์์์ ์ธ๊ธํ๋ฏ์ด Credit Wallet ๋ด์ ์ด๋ฒคํธ์ ์คํ๋ ๊ท๋ชจ์ ๋ณด์๊ท์ด ๊ฑธ๋ ค ์๋๋ฐ์. ๊ทธ๋งํผ ์ญ๋๊ธ ์ค์ผ์ผ์ด ๋ ์ ์๋๋ก ์ ๊ฒฝ์จ์ค ํ๊ตญ์ ์ ๋
์ด๋ฒคํธ ์ด๊ธฐ๋ ํ๋.. ๋ค ์ฑ๊ฒจ์ผ๊ฒ ์ง?
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
Be honest - how often do you check crypto prices?
30%Every hour
41%Multiple times daily
18%Once a day
11%Rarely/never
66 votes โข Final results ๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
Phone numbers were already tradeable assets.
666-6666 sold for ยฃ1.5 million.
867-5309 sold for $186,000+
Tokenization just made the market accessible.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
twitter was down and I had to talk to my family
never let this happen again @elonmusk
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
9 breakdowns. Three layers. One story: how @SentientAGI turns open models into something that actually learns and evolves.
This is the recap.
โฌโฌ
We started at the foundation. OML fingerprints every model. The Control Plane handles cryptographic governance. ROMA and GRID give open weights a real execution layer. Not just files. A system that runs.
Then we stress-tested it. LiveCodeBench Pro checking reasoning. SPIN-Bench testing strategy and memory. Every result feeds back into GRID so models don't just get scored, they improve.
Then it gets interesting. Intelligence stops being solo and starts being emergent. MindGames Arena running cooperation and deception. ROMA coordinating. GRID scaling it
โฌโฌ
Here's the full picture: OML feeds into benchmarks, benchmarks feed ROMA, ROMA scales through GRID, GRID powers MindGames, and it all loops back.
A self-improving intelligence system that you can verify and reward, built completely in public.
๐ข๐ ๐ + ๐ฅ๐ข๐ ๐ + ๐๐ฅ๐๐ โ ๐ง๐ต๐ฒ ๐๐๐น๐น-๐ฆ๐๐ฎ๐ฐ๐ธ ๐๐๐ ๐๐ผ๐ผ๐ฝ
Closing out this 9-post dive into @SentientAGI's architecture, from the model layer to the evaluation layer, and now the frontier where everything connects.
From open models to emergent intelligence, Sentient's architecture closes the loop, a living system where every model learns, adapts, and contributes back.
โต
๐ข๐ ๐ secures authorship and provenance. ๐๐ถ๐๐ฒ๐๐ผ๐ฑ๐ฒ๐๐ฒ๐ป๐ฐ๐ต and ๐ฆ๐ฃ๐๐ก-๐๐ฒ๐ป๐ฐ๐ต measure reasoning and problem-solving.
๐ ๐ถ๐ป๐ฑ๐๐ฎ๐บ๐ฒ๐ ๐๐ฟ๐ฒ๐ป๐ฎ trains coordination, deception, and social strategy.
๐ฅ๐ข๐ ๐ orchestrates the flow.
๐ง๐ต๐ฒ ๐๐ฅ๐๐ executes and connects everything in real time.
โต
Each layer feeds the next, proof, evaluation, emergence, until the network starts improving itself.
This is Sentient's endgame: an open AGI stack built by the community, not locked in corporate labs. Verifiable, rewardable, and alive. The foundation of a community-owned intelligence economy.
gSenti. Cheers๐
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐ข๐ ๐ + ๐ฅ๐ข๐ ๐ + ๐๐ฅ๐๐ โ ๐ง๐ต๐ฒ ๐๐๐น๐น-๐ฆ๐๐ฎ๐ฐ๐ธ ๐๐๐ ๐๐ผ๐ผ๐ฝ
Closing out this 9-post dive into @SentientAGI's architecture, from the model layer to the evaluation layer, and now the frontier where everything connects.
From open models to emergent intelligence, Sentient's architecture closes the loop, a living system where every model learns, adapts, and contributes back.
โต
๐ข๐ ๐ secures authorship and provenance. ๐๐ถ๐๐ฒ๐๐ผ๐ฑ๐ฒ๐๐ฒ๐ป๐ฐ๐ต and ๐ฆ๐ฃ๐๐ก-๐๐ฒ๐ป๐ฐ๐ต measure reasoning and problem-solving.
๐ ๐ถ๐ป๐ฑ๐๐ฎ๐บ๐ฒ๐ ๐๐ฟ๐ฒ๐ป๐ฎ trains coordination, deception, and social strategy.
๐ฅ๐ข๐ ๐ orchestrates the flow.
๐ง๐ต๐ฒ ๐๐ฅ๐๐ executes and connects everything in real time.
โต
Each layer feeds the next, proof, evaluation, emergence, until the network starts improving itself.
This is Sentient's endgame: an open AGI stack built by the community, not locked in corporate labs. Verifiable, rewardable, and alive. The foundation of a community-owned intelligence economy.
gSenti. Cheers๐
๐๐ฅ๐๐ ๐๐ป๐๐ฒ๐ด๐ฟ๐ฎ๐๐ถ๐ผ๐ป โ ๐ ๐๐น๐๐ถ-๐๐ด๐ฒ๐ป๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐ผ๐ป ๐ฆ๐ฒ๐ป๐๐ถ๐ฒ๐ป๐โ๐ ๐ฆ๐๐ฏ๐๐๐ฟ๐ฎ๐๐ฒ
gm. MindGames isn't running in some isolated sandbox, it's live on the GRID, @SentientAGI's substrate for open intelligence.
โง
Agents don't just simulate here. They exist in the same environment where OML models compete, interact, and evolve together.
Every action, message, and outcome gets logged directly to the GRID. OML tracks provenance. ROMA handles coordination. The GRID executes everything at scale. It's a closed feedback loop of learning and adaptation.
Each round of interaction refines the participants. Performance data loops back into training, turning competition into co-evolution.
โง
This is the architecture for collective learning, where Sentient's network becomes a living, self-optimizing intelligence web.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐๐ฅ๐๐ ๐๐ป๐๐ฒ๐ด๐ฟ๐ฎ๐๐ถ๐ผ๐ป โ ๐ ๐๐น๐๐ถ-๐๐ด๐ฒ๐ป๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐ถ๐ป๐ด ๐ผ๐ป ๐ฆ๐ฒ๐ป๐๐ถ๐ฒ๐ป๐โ๐ ๐ฆ๐๐ฏ๐๐๐ฟ๐ฎ๐๐ฒ
gm. MindGames isn't running in some isolated sandbox, it's live on the GRID, @SentientAGI's substrate for open intelligence.
โง
Agents don't just simulate here. They exist in the same environment where OML models compete, interact, and evolve together.
Every action, message, and outcome gets logged directly to the GRID. OML tracks provenance. ROMA handles coordination. The GRID executes everything at scale. It's a closed feedback loop of learning and adaptation.
Each round of interaction refines the participants. Performance data loops back into training, turning competition into co-evolution.
โง
This is the architecture for collective learning, where Sentient's network becomes a living, self-optimizing intelligence web.
๐ ๐ถ๐ป๐ฑ๐๐ฎ๐บ๐ฒ๐ ๐๐ฟ๐ฒ๐ป๐ฎ - ๐ช๐ต๐ฒ๐ฟ๐ฒ ๐๐ป๐๐ฒ๐น๐น๐ถ๐ด๐ฒ๐ป๐ฐ๐ฒ ๐๐ฒ๐๐ ๐ฆ๐ผ๐ฐ๐ถ๐ฎ๐น
Sentient's next move isn't just building smarter models, it's building models that learn through interaction.
gm. New series on @SentientAGI's Interaction Layer.
โง
MindGames Arena is the testbed for that. A multi-agent world where open models develop social reasoning: deception, cooperation, negotiation, coalition-building. All running in controlled, reproducible environments.
Built with researchers from Princeton, MIT, and Meta, it uses standardized, token-capped setups to test actual intelligence instead of just throwing more compute at problems.
โง
This is where coordination, adaptation, and strategy emerge in real time. You can watch collective behavior evolve when intelligence runs into uncertainty and has to figure things out.
MindGames is where open models learn to think together, the social layer of Sentient's AGI stack.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
one phone number connects your bank, exchange, social media, and shopping. see the problem, john?
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐ ๐ถ๐ป๐ฑ๐๐ฎ๐บ๐ฒ๐ ๐๐ฟ๐ฒ๐ป๐ฎ - ๐ช๐ต๐ฒ๐ฟ๐ฒ ๐๐ป๐๐ฒ๐น๐น๐ถ๐ด๐ฒ๐ป๐ฐ๐ฒ ๐๐ฒ๐๐ ๐ฆ๐ผ๐ฐ๐ถ๐ฎ๐น
Sentient's next move isn't just building smarter models, it's building models that learn through interaction.
gm. New series on @SentientAGI's Interaction Layer.
โง
MindGames Arena is the testbed for that. A multi-agent world where open models develop social reasoning: deception, cooperation, negotiation, coalition-building. All running in controlled, reproducible environments.
Built with researchers from Princeton, MIT, and Meta, it uses standardized, token-capped setups to test actual intelligence instead of just throwing more compute at problems.
โง
This is where coordination, adaptation, and strategy emerge in real time. You can watch collective behavior evolve when intelligence runs into uncertainty and has to figure things out.
MindGames is where open models learn to think together, the social layer of Sentient's AGI stack.
๐๐ผ๐ ๐๐ต๐ฒ ๐๐ฅ๐๐ ๐๐ฐ๐๐๐ฎ๐น๐น๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐
LiveCodeBench Pro and SPIN-Bench aren't just tests, they're feedback engines for @SentientAGI's GRID.
The GRID is where models, agents, and environments run together. Every time something executes, it generates signals: failures, successes, reasoning traces. Those don't disappear into some report. They loop back.
โง
Evaluation becomes training data. Performance becomes a signal for what to improve. The network learns from itself.
This is how Sentient scales differently. Open models get tested in transparent arenas, then reinforced and retrained through the GRID. Benchmarks turn into feedback that helps ROMA's agent layer coordinate improvements across the whole network.
โง
Instead of closed data locked in one company, the GRID turns evaluation into shared learning. A self-improving system owned by the community building it, not a single lab controlling it.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
You pay major carriers $80 monthly for a number tied permanently to your identity. Degenphone lets you own numbers without KYC for specific use cases.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐๐ผ๐ ๐๐ต๐ฒ ๐๐ฅ๐๐ ๐๐ฐ๐๐๐ฎ๐น๐น๐ ๐๐ฒ๐ฎ๐ฟ๐ป๐
LiveCodeBench Pro and SPIN-Bench aren't just tests, they're feedback engines for @SentientAGI's GRID.
The GRID is where models, agents, and environments run together. Every time something executes, it generates signals: failures, successes, reasoning traces. Those don't disappear into some report. They loop back.
โง
Evaluation becomes training data. Performance becomes a signal for what to improve. The network learns from itself.
This is how Sentient scales differently. Open models get tested in transparent arenas, then reinforced and retrained through the GRID. Benchmarks turn into feedback that helps ROMA's agent layer coordinate improvements across the whole network.
โง
Instead of closed data locked in one company, the GRID turns evaluation into shared learning. A self-improving system owned by the community building it, not a single lab controlling it.
๐ฆ๐ฃ๐๐ก-๐๐ฒ๐ป๐ฐ๐ต: ๐ง๐ฒ๐๐๐ถ๐ป๐ด ๐๐ต๐ฒ ๐๐ผ๐ป๐ด ๐๐ฎ๐บ๐ฒ ๐ผ๐ณ ๐ฅ๐ฒ๐ฎ๐๐ผ๐ป๐ถ๐ป๐ด
If LiveCodeBench Pro measures coding intelligence, SPIN-Bench tests something deeper: long-horizon, social, and strategic reasoning.
โง
Built with Princeton and UT Austin, SPIN-Bench combines four domains into one benchmark:
๐ฃ๐๐๐ ๐ฃ๐น๐ฎ๐ป๐ป๐ถ๐ป๐ด tests deep planning over large state spaces. ๐๐ผ๐ฎ๐ฟ๐ฑ ๐๐ฎ๐บ๐ฒ๐ measure adversarial reasoning under changing dynamics. ๐๐ผ-๐ผ๐ฝ ๐๐ฎ๐ฟ๐ฑ ๐๐ฎ๐บ๐ฒ๐ test coordination and inference under uncertainty. ๐ก๐ฒ๐ด๐ผ๐๐ถ๐ฎ๐๐ถ๐ผ๐ป ๐ง๐ฎ๐๐ธ๐ evaluate deception, coalition-building, and adaptive strategy.
Every agent loop follows the same pattern: observe, message, act. Same seeds, same budgets. Real strategy.
โง
@SentientAGI built SPIN-Bench to expose where current models still fail, brittle belief tracking, weak adaptation, shallow coordination, and to push the field toward agents that can actually think, not just respond.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
Some numbers are just special.
53 53 53 06 was one of them.
That user lost it to carrier bureaucracy.
With DegenPhone NFTs, it would've been protected.
We can't bring back 53 53 53 06.
But we can prevent the next legendary number from being lost.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐ฆ๐ฃ๐๐ก-๐๐ฒ๐ป๐ฐ๐ต: ๐ง๐ฒ๐๐๐ถ๐ป๐ด ๐๐ต๐ฒ ๐๐ผ๐ป๐ด ๐๐ฎ๐บ๐ฒ ๐ผ๐ณ ๐ฅ๐ฒ๐ฎ๐๐ผ๐ป๐ถ๐ป๐ด
If LiveCodeBench Pro measures coding intelligence, SPIN-Bench tests something deeper: long-horizon, social, and strategic reasoning.
โง
Built with Princeton and UT Austin, SPIN-Bench combines four domains into one benchmark:
๐ฃ๐๐๐ ๐ฃ๐น๐ฎ๐ป๐ป๐ถ๐ป๐ด tests deep planning over large state spaces. ๐๐ผ๐ฎ๐ฟ๐ฑ ๐๐ฎ๐บ๐ฒ๐ measure adversarial reasoning under changing dynamics. ๐๐ผ-๐ผ๐ฝ ๐๐ฎ๐ฟ๐ฑ ๐๐ฎ๐บ๐ฒ๐ test coordination and inference under uncertainty. ๐ก๐ฒ๐ด๐ผ๐๐ถ๐ฎ๐๐ถ๐ผ๐ป ๐ง๐ฎ๐๐ธ๐ evaluate deception, coalition-building, and adaptive strategy.
Every agent loop follows the same pattern: observe, message, act. Same seeds, same budgets. Real strategy.
โง
@SentientAGI built SPIN-Bench to expose where current models still fail, brittle belief tracking, weak adaptation, shallow coordination, and to push the field toward agents that can actually think, not just respond.
๐๐ถ๐๐ฒ๐๐ผ๐ฑ๐ฒ๐๐ฒ๐ป๐ฐ๐ต ๐ฃ๐ฟ๐ผ: ๐๐ฒ๐ป๐ฐ๐ต๐บ๐ฎ๐ฟ๐ธ๐ถ๐ป๐ด ๐๐ผ๐ฑ๐ถ๐ป๐ด ๐๐ป๐๐ฒ๐น๐น๐ถ๐ด๐ฒ๐ป๐ฐ๐ฒ ๐ง๐ฟ๐ฎ๐ป๐๐ฝ๐ฎ๐ฟ๐ฒ๐ป๐๐น๐
Gm, stepping into the Evaluation Layer where open models meet open measurement.
Most coding benchmarks test recall.
@SentientAGI built one that tests reasoning.
โง
๐๐ถ๐๐ฒ๐๐ผ๐ฑ๐ฒ๐๐ฒ๐ป๐ฐ๐ต ๐ฃ๐ฟ๐ผ runs code inside controlled sandboxes, tracking compilation, execution, and runtime feedback in real time.
Every output is verified automatically.
Every result is reproducible.
Every dataset is open.
No private scoring. No hidden leaderboards. Just transparent, scientific evaluation that rewards actual intelligence, not memorization.
โง
For Sentient, this isnโt just benchmarking, itโs accountability for the open model era.
Because if intelligence is going to be open, the way we measure it must be open too.
Transparent evaluation โ reproducible progress โ a stronger collective intelligence.
Thatโs how Sentientโs evaluation layer keeps the loop honest.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
how many services have your real, primary phone number? what happens when one leaks it?
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
Tokenized numbers already exist. Virtual numbers that work within specific platforms have been NFTs for a while now. Some work only for specific apps like fragmentโs anonymous number works for telegram.
Degenphone is different because we tokenized actual eSIMs.
Virtual numbers are software identifiers that route messages within closed systems. eSIMs are real telecommunications infrastructure. One works where the platform allows and the other works wherever your regular phone number would.
Our numbers receive SMS from any service right now. Banking apps, exchange verification, social media. If it sends a text to a phone number, it works with Degenphone. Voice calls implementation is not far away.
We're even exploring direct installation on phones. Pop the NFT's eSIM profile into your device like any carrier eSIM. Just your phone working normally, except you own the number.
Virtual number NFTs proved people want to own their communications. We're proving they can own real phone infrastructure.
Other projects tokenized virtual numbers. We tokenized actual phone numbers.
That's why Degenphone works everywhere, not somewhere.
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
Over $1B in leveraged longs got liquidated in the past 24 hours.
This entire sell off was pure manipulation.
Theyโre shaking out weak hands before the next leg up or before the heavy bear.
Stay strong.
Whatโs the worst that can happen .
Best to stay locked in on @HeyElsaAI
Might even take a dip on katana
๐ฐ๐ฒ๐ฐ๐ธ๐ต๐ธ retweeted
๐ข๐ ๐ ๐ช๐ผ๐ฟ๐ธ๐๐ต๐ผ๐ฝ: ๐๐ฟ๐๐ฝ๐๐ผ๐ด๐ฟ๐ฎ๐ฝ๐ต๐ถ๐ฐ ๐๐ผ๐ป๐๐ฟ๐ผ๐น ๐ฃ๐น๐ฎ๐ป๐ฒ ๐ณ๐ผ๐ฟ ๐ข๐ฝ๐ฒ๐ป ๐ ๐ผ๐ฑ๐ฒ๐น๐
Gm gm, still on the Model Layer.
In the last post, we talked about fingerprints. Todayโs about what happens when those fingerprints become policies.
โ
Most open models today are like unlocked safes, once the weights are out, anyone can do whatever they want with them.
@SentientAGI's OML Workshop Track fixes that by adding a cryptographic control layer on top of the model. Think of it as a security system that governs how open weights get used, without locking them behind some API.
๐ง๐ต๐ฒ ๐ฑ๐ฎ๐๐ฎ ๐ฝ๐น๐ฎ๐ป๐ฒ handles normal token generation.
๐ง๐ต๐ฒ ๐ฐ๐ผ๐ป๐๐ฟ๐ผ๐น ๐ฝ๐น๐ฎ๐ป๐ฒ manages keys, policy rules, and signed records of who ran what and where.
โ
Together, they make open models both accountable and policy-aware. Things like academic-only use, region restrictions, or commercial permissions get enforced automatically.
With OML, openness doesn't mean chaos anymore, it means verifiable control for the builders powering Sentient's intelligence layer.
๐ข๐ ๐ 1.0: ๐ฆ๐ฐ๐ฎ๐น๐ฎ๐ฏ๐น๐ฒ ๐๐ถ๐ป๐ด๐ฒ๐ฟ๐ฝ๐ฟ๐ถ๐ป๐๐ถ๐ป๐ด ๐ณ๐ผ๐ฟ ๐ข๐ฝ๐ฒ๐ป ๐ ๐ผ๐ฑ๐ฒ๐น๐
Yooo, new series, stay with me๐
@SentientAGI just changed how open-model provenance works.
OML 1.0 introduces scalable fingerprinting, embedding 24,576 invisible identifiers into LLMs without touching performance. For context, the previous record was around 100๐
โ
These fingerprints stick around through fine-tuning, distillation, even model merges. Every open weight can be traced back to whoever created it, even when the model's running locally on someone's machine.
First time open-source intelligence can be both shareable and accountable at the same time. Builders don't give up ownership when they open their work, they get proof instead.
โ
This is what Sentient means by open, monetizable, loyal models.
Open-source doesn't have to mean free-for-all anymore. OML makes it actually sustainable. gSenti๐ซถ๐ฝ
On the next series, we'll talk
OML Workshop: Cryptographic Control Plane for Open Models