Entropy deploy program
К концу этого руководства вы развернете программу на блокчейне Entropy и сможете взаимодействовать с ней для создания подписей.

A project is only as strong as its team. Meet the founders of Gensyn, who combine deep AI research w…
A project is only as strong as its team. Meet the founders of @gensynai, @fenbielding & @_grieve, who combine deep AI research with a crypto-cypherpunk ethos.

The Bull Case for @gensynai: A $50M Bet on the Future of Intelligence.
Why did @a16zcrypto lead a $43M round into @gensynai? It wasn’t just about GPUs. It was a bet on who will own the infrastructure for intelligence itself. Let’s break down the bull case.
<100 subscribers
Entropy deploy program
К концу этого руководства вы развернете программу на блокчейне Entropy и сможете взаимодействовать с ней для создания подписей.

A project is only as strong as its team. Meet the founders of Gensyn, who combine deep AI research w…
A project is only as strong as its team. Meet the founders of @gensynai, @fenbielding & @_grieve, who combine deep AI research with a crypto-cypherpunk ethos.

The Bull Case for @gensynai: A $50M Bet on the Future of Intelligence.
Why did @a16zcrypto lead a $43M round into @gensynai? It wasn’t just about GPUs. It was a bet on who will own the infrastructure for intelligence itself. Let’s break down the bull case.
Share Dialog
Share Dialog


🧠 How does Gensyn guarantee AI work is done correctly across 1000s of untrusted devices? The answer is a brilliant system called Verde. It's the project's biggest moat and the key to unlocking a truly permissionless AI network.
Here's how it works 👇
The Problem: The Chaos of Different Hardware
Verifying decentralized AI training is incredibly hard. Why? Because different GPUs (even from the same manufacturer) can produce slightly different results for the same task due to tiny architectural variations. This is called non-determinism. In a trustless network, if two honest nodes get different answers, how do you know who is right? You can't build a reliable system on inconsistent results.
Step 1: RepOps for Perfect Consistency
@gensynai's first innovation is Reproducible Operators (RepOps). Think of RepOps as a universal translator for hardware. It's a library that forces every device — from an NVIDIA A100 to a gaming GPU — to perform calculations in the exact same order, ensuring they produce bitwise identical outputs. This creates a single source of truth. Now, if two nodes disagree, one of them must be incorrect.
Step 2: The Ultra-Efficient Dispute Game
When a disagreement occurs, re-running the entire training job to find the error would be incredibly expensive and slow. This is where Verde's "refereed delegation" shines. Instead of re-doing everything, Verde initiates a hyper-efficient dispute resolution game to pinpoint the exact moment of divergence.
- Phase 1: The protocol rapidly identifies the specific training step where the two nodes' calculations first differed.
- Phase 2: It then drills down further, identifying the single mathematical operator (e.g., one matrix multiplication) within that step that caused the error. A neutral referee (like a smart contract) only needs to re-compute that one tiny operation to determine the honest party.
Why It's Better Than ZK-Proofs (For This Job)
Cryptographic proofs like ZK-proofs are powerful but prohibitively expensive for verifying massive-scale ML workloads. Verde's approach is dramatically more efficient — by several orders of magnitude. It provides robust security guarantees (assuming at least one verifier is honest) without the crippling computational overhead, making decentralized AI training economically viable.
The Result: Scalable, Permissionless Verification
Verde is @gensynai's secret weapon. It solves the fundamental challenge of trust in a decentralized compute network. By making verification cheap, fast, and reliable, Verde is the core building block that allows @gensynai to permissionlessly onboard every compute device in the world, creating the scalable, open network needed to democratize artificial intelligence.
Follow me, and I will tell you in detail everything you need to know about this project — from its groundbreaking technology to its role in the future of intelligence.
Stay tuned!
#Gensyn #AI #DePIN #Crypto
🧠 How does Gensyn guarantee AI work is done correctly across 1000s of untrusted devices? The answer is a brilliant system called Verde. It's the project's biggest moat and the key to unlocking a truly permissionless AI network.
Here's how it works 👇
The Problem: The Chaos of Different Hardware
Verifying decentralized AI training is incredibly hard. Why? Because different GPUs (even from the same manufacturer) can produce slightly different results for the same task due to tiny architectural variations. This is called non-determinism. In a trustless network, if two honest nodes get different answers, how do you know who is right? You can't build a reliable system on inconsistent results.
Step 1: RepOps for Perfect Consistency
@gensynai's first innovation is Reproducible Operators (RepOps). Think of RepOps as a universal translator for hardware. It's a library that forces every device — from an NVIDIA A100 to a gaming GPU — to perform calculations in the exact same order, ensuring they produce bitwise identical outputs. This creates a single source of truth. Now, if two nodes disagree, one of them must be incorrect.
Step 2: The Ultra-Efficient Dispute Game
When a disagreement occurs, re-running the entire training job to find the error would be incredibly expensive and slow. This is where Verde's "refereed delegation" shines. Instead of re-doing everything, Verde initiates a hyper-efficient dispute resolution game to pinpoint the exact moment of divergence.
- Phase 1: The protocol rapidly identifies the specific training step where the two nodes' calculations first differed.
- Phase 2: It then drills down further, identifying the single mathematical operator (e.g., one matrix multiplication) within that step that caused the error. A neutral referee (like a smart contract) only needs to re-compute that one tiny operation to determine the honest party.
Why It's Better Than ZK-Proofs (For This Job)
Cryptographic proofs like ZK-proofs are powerful but prohibitively expensive for verifying massive-scale ML workloads. Verde's approach is dramatically more efficient — by several orders of magnitude. It provides robust security guarantees (assuming at least one verifier is honest) without the crippling computational overhead, making decentralized AI training economically viable.
The Result: Scalable, Permissionless Verification
Verde is @gensynai's secret weapon. It solves the fundamental challenge of trust in a decentralized compute network. By making verification cheap, fast, and reliable, Verde is the core building block that allows @gensynai to permissionlessly onboard every compute device in the world, creating the scalable, open network needed to democratize artificial intelligence.
Follow me, and I will tell you in detail everything you need to know about this project — from its groundbreaking technology to its role in the future of intelligence.
Stay tuned!
#Gensyn #AI #DePIN #Crypto
No comments yet