You type "Convert 0.1 ETH to USDC at the best rate" and the agent takes it from there, finds the best route across Uniswap V3 fee tiers, checks policy limits, signs through the OWS vault and broadcasts a real swap on Sepolia. The private key never touches the agent. All signing stays inside the OWS vault. My intent-to-action agent demo built for the Open Wallet Standard Hackathon. Check out the video to see it in action 👇 🔗 github.com/crazywriter1/ows-… @OpenWalletStd @OpenWallet

Dünya genelinde kripto merkezi konumuna gelmiş ülkelerin mevcut koşullarda rekabet gücünü kaybettiğini (savaş vb kondisyonlardan ötürü), bu alanda ağır yaptırımlar (yüksek vergilendirme, yasaklama vb.) uygulayan ülkelerin ise söz konusu merkezlere sermaye kaçışını önlemek amacıyla adım adım geri çekilmeye başladığını net biçimde gözlemliyoruz. Bu tablo bize önemli bir fırsat penceresi sunmaktadır: Dünyaya öncü olabilecek, cazip ve yenilikçi bir düzenleyici çerçeve oluşturarak uluslararası ilgiyi Türkiye'ye yönlendirmek ve ülkemize kalıcı döviz akışı sağlamak artık bir tercih değil, stratejik bir zorunluluktur. Bu kazanım yalnızca vergi gelirleriyle sınırlı kalmayacak; gayrimenkul işlemleri, turizm ve yeni işletmelerin ülkemizde kurulmasıyla birlikte dolaylı katkıların toplamı son derece ciddi bir ekonomik atılıma ve kalıcı katma değere dönüşecektir. Aksi takdirde uygulanan kısıtlayıcı tedbirler; mevcut kurumların işlevsiz hale gelmesine, sermayenin OTC ve P2P kanalları aracılığıyla yurt dışına akmasına ve kayıt dışılığın fiilen derinleşmesine zemin hazırlayacaktır. Bu sonuç, ne hazine ne de sektör için kabul edilebilir bir senaryo değildir. Sayın Bakanım, yasa tasarısı düzenlenirken yalnızca anlık vergi geliri hesaplamaları değil, akabinde devreye girecek dolaylı etki zincirinin tamamı göz önünde bulundurulmalıdır. Konuya bu bütünsel perspektifle yaklaşılacağına olan inancımı koruyarak, adil ve dengeli bir düzenlemenin ortaya çıkmasını temenni ediyorum.

Sending files shouldn’t mean trusting a random server with your data. So I built Shelby Send. → Encrypts files in your browser → Optional password protection → Share with a simple link Your machine encrypts, they decrypt. that’s it. Live: shelby-send.vercel.app Code: github.com/crazywriter1/shel… Feedback welcome. @shelbyserves

Built RugRadar - a Telegram bot powered by Hermes Agent by @NousResearch 🤖 Send a token address and it instantly returns a rugpull risk report. Checks onchain signals and market data to help spot suspicious tokens faster. Stack: Python, Telegram Bot API, Hermes Agent, Dexscreener API, Etherscan API, etc. Still learning and improving it but it's open source 👇 github.com/crazywriter1/Herm…

I built a provably fair VRF playground using @commonwarexyz cryptography. 🎲 Dice 🪙 Coin Flip 🍀 Lottery Try it: commonware-vrf-dice-producti… Every result includes a verifiable VRF proof, anyone can independently verify it. Built with Ed25519 SHA-256, featuring player identity, leaderboard tracking and shareable proof verification. Code: github.com/crazywriter1/comm… #Commonware

I just turned a classic game into an onchain mini app on @base Checkers 𖣯🔴 Finish a match and you can submit your score onchain through a smart contract (small fee). No setup. No signups. Just open and play. Play here: • Base App → base.app/app/checkers-ebon.v… • Farcaster Mini App → farcaster.xyz/miniapps/LQKaP… Built as a small experiment to explore mini apps onchain interactions. Bringing a simple game directly into social and onchain apps. Huge thanks to @UfukDegen for the guide 🙌 Feedback welcome!

Privacy shouldn’t be optional on blockchain. @zama enables confidential applications on any blockchain using Fully Homomorphic Encryption (FHE), allowing data to be processed while remaining encrypted without sacrificing verifiability. A foundation for private onchain computation.

Announcing SERA-Crypto (Semantic Embedding & Reasoning Agent): our new reasoning architecture built for SOTA crypto research. #1 open-source agent on DMind #1 on our live crypto benchmark Outperforms GPT-5, Grok 4, Gemini 2.5 Pro, and Perplexity Finance…all under 45 seconds.

Seeing $SENT show up on @Coinbase’s radar feels like another big step forward. Not because of hype but because it shows how far this open source AGI movement has already come. From ROMA and ODS to the rapidly growing GRID ecosystem. @SentientAGI is finally getting the attention its builders and community earned through real work. Whatever comes next, one thing is clear: This momentum is only getting stronger.

Diverse Expert Ensembles are built on a simple idea: No single model sees everything. Each expert has different strengths, different biases and different ways of interpreting data. @gensynai research shows that when these experts are combined, the system becomes more stable, more balanced and more reliable across a wide range of inputs. Instead of depending on one model’s blind spots, an ensemble can use multiple perspectives to reach stronger predictions. It’s a reminder that in machine learning. Just like in real life. Diversity is not a complication. It’s an advantage.

Sentient is often introduced through tools and agents but at its core it’s something much deeper. An open source research lab focused on advancing reasoning itself. From ROMA’s recursive agent architecture to ODS’s deep research engine, from model fingerprinting to agentic security, every piece of work is centered around one idea: Building AI systems that can think clearly, verify their actions and operate safely in real world, high stakes environments. GRID then brings this research into the world. Uniting developers, researchers, and enterprises under one ecosystem where open source intelligence can actually grow and coordinate. As @SentientAGI continues sharing its findings at events like NeurIPS, the goal remains the same: Open research, open collaboration, open intelligence.

⚡ Gensyn’s research stack tackles two major challenges in large scale ML training: Resilience and communication cost. 🔹CheckFree introduces a fault tolerant approach that lets training continue even if a node fails without having to restart from scratch. 🔸SkipPipe reduces unnecessary data transfers in pipeline parallel models, helping lower iteration times and making deep pipelines more efficient. ▫️Together, they address two real bottlenecks in modern ML: Interruptions and communication overhead. These kinds of research efforts are a core part of how @gensynai is building a scalable and reliable compute infrastructure.

Airdrop allocations are still cooking 🪂 Contributions are still live and count toward your final airdrop allocation until TGE 👀 PS: Higher mindshare = higher rewards.

Gensyn’s reward model is built on a simple principle: You don’t earn for turning a node on. You earn for producing real, verifiable ML compute. Every task runs on external hardware, every result is checked through systems like Verde and a payout only happens when the network can prove the computation was executed correctly. No shortcuts, no blind trust, no hidden rules. This 'execution → verification → payment' flow is what keeps the economy fair. It means contributors get rewarded for actual work and the network never pays for fabricated or inconsistent outputs. By aligning incentives with verifiable compute, @gensynai creates an open and transparent economy where anyone can contribute and everyone is rewarded based on what they actually deliver. Nothing more, nothing less.

Gensyn’s testnet runs on a mix of node types, each with a clear job. GPU nodes take on the heavy training steps, CPU nodes handle lighter compute and coordination and Verifier nodes (powered by systems like Verde and Judge) make sure every result is valid. It’s a simple but powerful design: different hardware working together under a single trust layer. This is how @gensynai turns decentralized compute into something reliable, scalable and actually usable.

Centralized AI vs Decentralized AI: It’s not about cost, it’s about trust. In centralized systems, you send an AI task, you get a result and everything in between is a black box. ❓Who computed it? ❓Was it done correctly? ❓Was the model really trained the way they claim? You don’t know, you just trust. That’s exactly where @gensynai changes the game. Gensyn distributes AI compute across thousands of independent nodes and makes every step verifiable. Tasks are broken down, computations are cryptographically signed and results come with proofs not promises. The old I hope this output is correct era is over. So the real issue isn’t cost. It’s trust. And Gensyn delivers exactly that: ✅ Transparency ✅ Verifiable computation ✅ An open network where anyone can contribute ✅ Proof backed AI instead of blind belief The future of AI isn’t just about bigger models. It’s about building an ecosystem you can actually trust and that’s what Gensyn is here for.

Verde is one of those systems you don’t really see but everything quietly depends on it. No matter where a training job runs; on a high end server, a small home GPU or anything in between. Verde steps in to check whether every result is actually correct. No guessing. No blind trust. No hoping the node did the right thing. And honestly, that’s what makes @gensynai feel different. A decentralized compute network can only work if every contribution is verifiable and Verde turns that idea into reality. It keeps the whole system honest, consistent and transparent, even as the network grows and contributors become more diverse. In a space full of trust us, Verde brings something refreshingly rare: proof.

CodeZero is one of the most underrated pieces of the @gensynai testnet and it quietly does the heavy lifting. It takes worker performance, verification results and task outcomes… And turns them into a clear, consistent reputation score. No magic. No hidden formulas. Just transparent metrics that decide who gets more work, who gets paid fairly and who needs to improve. A healthy network needs healthy signals and CodeZero is exactly that: The heartbeat that keeps the ecosystem honest and competitive.