Introduce EAGLE, a new method for fast LLM decoding... based on compression: - 3x🚀than vanilla - 2x🚀 than Lookahead (on its benchmark) - 1.6x🚀 than Medusa (on its benchmark) - provably maintains text distribution - trainable (in 1~2 days) and testable on RTX 3090s Playground: Blog: Code: ⚒️First Principle: Compression! Yi Ma We find that the sequence of second-top-layer features is compressible, making the prediction of subsequent feature vectors from previous ones easy by a small model. 🙏Acknowledge: This project is greatly inspired by the Medusa team (Tianle Cai ✈️ ICML🇰🇷 @yli3521 Zhengyang Geng ✈️ ICML Hongwu Peng Tri Dao), the Lookahead team (Hao Zhang LMSYS Org), and others. Joint work with Yuhui Li and Chao Zhangshow more

Jointly announcing EAGLE-3 with SGLang: Setting a new record... in LLM inference acceleration! - 5x🚀than vanilla (on HF) - 1.4x🚀than EAGLE-2 (on HF) - A record of ~400 TPS on LLama 3.1 8B with a single H100 (on SGLang) - 1.65x🚀in latency even for large bs=64 (on SGLang) - A new scaling law: more training data, better speedup - Apache 2.0 Paper: Code: SGLang version: ⚒️Takeaway: Introducing training-time test, a novel draft model training technique: we replace feature prediction with direct token prediction and shift from top-layer-only features to multi-layer feature fusion. This approach unlocks a new scaling law previously undiscovered in EAGLE and EAGLE-2. 🙏Acknowledge: We would like to thank the SGLang team (zhyncs Lianmin Zheng Ying Sheng James Liu, Ke Bao, and others LMSYS Org) for their merge and careful evaluation of EAGLE-3 on SGLang. 🤝Want to collaborate? We're a small academic group with limited GPU resources. If you're interested in supporting our next version of EAGLE or would like us to train a preliminary version tailored to a specific model, please get in touch! Joint work with Yuhui Li, Fangyun Wei, and Chao Zhangshow more

Prediction Market is one of the leading Web3 niches... in 2024! With about $4 Billion in trading volume, $192 Million in TVL and millions of users in 2024, prediction protocols have been showing tremendous growth and attracting user adoption in Web3. PolyMarket seems to be leading the pack, with over $175M in TVL, and backing from Vitalik Buterin. However, a majority of Prediction Markets, including PolyMarket, currently lacks the flexibility and capital efficiency required for seamless transactions. Also, they are all majorly focused on driving Web2 users thus neglecting the need for markets that cater for short-term, high-risk investments from Web3 Degens. To tackle these flaws, there's a need for a revolutionary contender that understands the need of Web3 Chads That's where Predict Hub comes in PredictHub is a prediction Market that transforms real world events into opportunities for everyone to participate and forecast. Launching on Arbitrum, PredictHub is already catching the attention of major players in the Space by offering something PolyMarket and others don't - Flexibility and Incentives. By offering fast market updates and innovative prediction category like ETF Forecasts, PredictHub is changing how we interact with Prediction Markets. But then, here's where it gets more interesting; PredictHub offer users a unique point system, where you don't just make predictions, you also earn rewards. The more you Predict, the more you earn. These rewards are 2-fold: Nova and Orbit Points. Nova Points are earned by traders based on their trading activity and their leaderboard ranking. Orbit Points, on the other hand, are earned by users who provide liquidity, based on their LP size and duration. Other Point systems include PolyMarket User Points, Leaderboard Bonus and Market Multipliers. These rewards offer users more competitive edge than other prediction markets. Apart from these rewards, PredictHub features a unique 3-tier referral system, rewarding users with even more as you invite your friends. The more friends you bring, the greater the rewards. On top of these, PredictHub focuses on USDC and a wide-range of yield bearing assets like GLP, gUSDC, and sUSDe, enabling users to optimise their earning while holding assets across Networks. Exciting, right? PredictHub is in its Testnet phase and you can start earning Points Right away 🔅 Here's how to Get Started on PredictHub: 1. Go to 2. Request Faucet 3. Start making predictions and earning Points Easy-Peasy ✅ More Info can be gotten from Predict Hub All eyes are on PredictHub as the fix for the flaws of Prediction Market Protocols. With its unique approach targeting untapped niches that most existing prediction markets have yet to explore, I believe the Protocol has the potential to become a breakout success I will be placing good Predictions to Position 🚀🚀🚀show more

Model-Free Reinforcement Learning (MFRL) has been alluring, especially with... supercharged compute with physics on GPU. However, the methods use 0-th order gradients, and are often not the best optimizers. Can we do better than PPO in continuous control for robotics? Turns out yes! 🥳 tl;dr: Faster, better RL than PPO in continuous control 💪 The answer lies in using more information from the simulation. We are juicing the simulation on GPU as it is, why not use it for gradients as well? This has been a driving question in a series of our works. We first studied this problem in ICLR 2022 paper on Short Horizon Actor Critic Naive gradient based methods are stuck in local minima and have exploding/vanishing gradients. SHAC solved this problem truncated rollouts and model based value estimation, where the model is Differentiable Sim. This boosted sample efficiency and wall-clock time immensely especially in high dimensional systems such as humanoids Yet, given enough compute PPO often caught up. Our follow up paper on on Adaptive Horizon Actor Critic at ICML 2024 discovers the cause and provides a fix. However, we find that even when given ground-truth dynamics, not all gradients are useful due to sample error. 1st-Order Model-Based Reinforcement Learning methods employing differentiable simulation provide gradients with reduced variance but are susceptible to bias in scenarios involving stiff dynamics, such as physical contact. We find that back-propagating through contact and long trajectories drastically reduces gradient accuracy. Using this insight, we propose AHAC to dynamically adapt its roll-out horizon to avoid differentiating through stiff contact. AHAC is a first-order model-based RL algorithm that learns high-dimensional tasks in minutes (wall clock) and outperforms PPO by 40%, even in the limit of data provided to PPO. This work is led by Ignat Georgiev alongside Krishnan Srinivasan, Jie Xu, Eric Heiden and ample assistance from warp team at NVIDIA Robotics (Miles Macklin)show more

We’re excited to introduce ShinkaEvolve: An open-source framework that... evolves programs for scientific discovery with unprecedented sample-efficiency. Blog: Code: Like AlphaEvolve and its variants, our framework leverages LLMs to find state-of-the-art solutions to complex problems, but using orders of magnitude fewer resources! Many evolutionary AI systems are powerful but act like brute-force engines, burning thousands of samples to find good solutions. This makes discovery slow and expensive. We took inspiration from the efficiency of nature. ‘Shinka’ (進化) is Japanese for evolution, and we designed our system to be just as resourceful. On the classic circle packing optimization problem, ShinkaEvolve discovered a new state-of-the-art solution using only 150 samples. This is a big leap in efficiency compared to previous methods that required thousands of evaluations. We applied ShinkaEvolve to a diverse set of hard problems with real-world applications: 1/ AIME Math Reasoning: It evolved sophisticated agentic scaffolds that significantly outperform strong baselines, discovering an entire Pareto frontier of solutions trading performance for efficiency. 2/ Competitive Programming: On ALE-Bench (a benchmark for NP-Hard optimization problems), ShinkaEvolve took the best existing agent's solutions and improved them, turning a 5th place solution on one task into a 2nd place leaderboard rank in a competitive programming competition. 3/ LLM Training: We even turned ShinkaEvolve inward to improve LLMs themselves. It tackled the open challenge of designing load balancing losses for Mixture-of-Experts (MoE) models. It discovered a novel loss function that leads to better expert specialization and consistently improves model performance and perplexity. ShinkaEvolve achieves its remarkable sample-efficiency through three key innovations that work together: (1) an adaptive parent sampling strategy to balance exploration and exploitation, (2) novelty-based rejection filtering to avoid redundant work, and (3) a bandit-based LLM ensemble that dynamically picks the best model for the job. By making ShinkaEvolve open-source and highly sample-efficient, our goal is to democratize access to advanced, open-ended discovery tools. Our vision for ShinkaEvolve is to be an easy-to-use companion tool to help scientists and engineers with their daily work. We believe that building more efficient, nature-inspired systems is key to unlocking the future of AI-driven scientific research. We are excited to see what the community builds with it! Learn more in our technical report:show more

Building The On-Chain Cooperative 🟡 Welcome to the dawn... of a new era in the crypto space, where the buzzword "community" is not just a hollow echo but a vibrant force that propels us towards a brighter future. Let's delve into the heart of MODE, the Onchain Cooperative that seeks to redefine the landscape of web3. What does MODE stand for? MODE stands for building an on-chain cooperative focused on sustainable growth and collective prosperity. At its core, MODE is guided by the principles of cooperation, shared incentives, and community-driven development. The goal is to shift from the "fat protocol" mentality where most value accrues to the blockchain/protocol itself, towards an ecosystem where builders, users, and applications can thrive together. What’s MODE's vision and mission in the web3 space? MODE's vision is to return to web3's founding promise - a future that is better for all, not just the individual. A world with aligned incentives that drive growth for everyone involved. A place with opportunities for all, not just the few. The mission is to pioneer the on-chain cooperative - where contributors are rewarded fairly based on the value they provide. Features like Sequencer Fee Sharing distribute a portion of fees to smart contract developers, incentivizing participation. The aim is to encourage collaboration instead of confrontation. Together, the MODE community can deliver new models for cooperation and shared prosperity in web3. Mode Network will solve many problems today in Web3: • Lack of incentives for developers: Developers creating decentralized apps (dApps) currently have few direct economic incentives to create and maintain their projects. Mode provides them with a steady source of income through fee-sharing. • Lack of collaboration: There are few incentives for blockchain projects to compete less and collaborate more for the benefit of the entire ecosystem. Mode's model encourages collaboration by aligning participants economically. • Excessive value accrual at the protocol layer: Mode aims for a more balanced model where the protocol's success is fueled by the success of application developers/builders and the wider community. Growth is a two-way street – "as we grow, you grow". The MODE Pledge 💛 The promise of crypto and blockchain is a brighter future. One that is better for all not just the individual. Where nothing is more important than community. We've strayed from this path. Entering a world of player vs player. Where value is extracted rather than shared. The game is zero sum rather than positive sum. And incentives are aligned with domination, rather than cooperation. Mode is the dawn of a new age. and a return to the promise of what can be. A world with aligned incentives that drive growth for builders, users and projects. A place with opportunities for all, rather than the few. Where we say goodbye to the 'fat protocol', and hello to the onchain cooperative. Join us on our mission to grow together. If this vision for a community-powered web3 ecosystem resonates - where creators are rewarded for their contributions - you can join the MODE on-chain cooperative! Visit Join the discord community Follow Mode 🟡 Together, we can transform web3 into a positive-sum game that unlocks new possibilities for all. Where your growth fuels the growth of others. Let's build the on-chain cooperative!show more

Don't train the model, evolve the harness. I read... a brilliant blog post from Hugging Face where they took a frozen open model scoring 0% on a hard legal agent benchmark, left its weights alone, and let an automated loop rewrite only the code around it. That code layer is the harness, the runtime wrapper that feeds the model context, runs its tool calls, and decides when a run ends. By the time the loop finished, the system had essentially matched Sonnet 4.6 on the benchmark's headline metric, at roughly 7x lower cost per task. Zero weights changed. The gain existed because of where the model was failing. The judge only grades files saved in the right place under the exact requested filename, and the model kept doing the legal analysis correctly, then saving it under the wrong name, dropping it in a scratch folder, or never writing it at all. So the 0% was never measuring legal reasoning. It was measuring the harness. Hand-tuning that layer is slow and model-specific, so they automated it. A Claude proposer adds exactly one mechanism per iteration, and an outer loop keeps it only if it clearly beats the current best, so accepted mechanisms compound. What the loop discovered says a lot about where agents actually fail. → The biggest single gain was file handling, not intelligence. An automatic step that lands the deliverable exactly where the judge expects it beat every prompt change, with zero extra model tokens. → Code fixes transferred across models, prompt playbooks did not. The same harness lifted a smaller model from the same family by 14 points, but the tuned prompts hurt a different model family on tasks it could already finish. → The harness mattered more than anything else. Same model, same judge, same tasks, and five different harnesses scored anywhere between 3.5% and 80.1%. The gains do eventually flatten, and the remaining misses look like real capability gaps. At some point the wrapper runs out of tricks and the model has to carry the work. But the lesson holds. A benchmark score measures the model and its harness together, and until the harness is fixed, it's impossible to know which one failed. I highly recommend reading this: I also wrote a deep dive on agent harness engineering a while back, covering the orchestration loop, tools, memory, context management, and everything that turns a stateless LLM into a capable agent. The article is quoted below.show more

Had an awesome time with Mark Haywood and the... $edm.v team on the site tour! The DFO permit is a near term major catalyst, which upon receiving will be followed by the final production decision. The water at the polishing pond after the tailings pond is astoundingly clean and loaded with fish, and tested as much cleaner than the water of the local streams and rivers. DFO permits are very rarely not granted, and $EDM.v's thoroughness and hard work is sure to be rewarded. The provincial government of Nova Scotia is also very supportive of the project and is eager to see them enter production. EDM's cash flow for next year will likely be greater than the current mcap, only to continue ramping up to double or greater than the current mcap. Very few juniors are this close to cash flow with such a small capex needed for restart, which will be financed with debt and paid back in around half a year. $edm.v should be 3-4x higher now, at 0.5x NPV, and even higher once the DFO permit is received this summer, 0.7-0.8x NPV. Right now we are trading near 0.10-0.15x NPV when you adjust for the DMS update, current metal prices, as well as doubled Gypsum prices, which is crazy cheap for how advanced the project is. Another impression from the tour is how solid and well made the 2700tpd mill is. It was built by Imperial Oil, a multi billion dollar company, who put in a lot of extras. There are about 5-10m CAD in replacement and maintainence parts on site from the previous opperators, still in original packaging. All of this is just the base case projection, it doesn't include the $gold optionality. Once the mine is in production and the main put is drained, they can go back to where the last producers will pulling ore and resuming mining high grade gold, which left off at 20k to 60k ounces a year based on a rough back calculation from the left over concentrates. This gold is proven via production and recovery, its absolutely there, and the team is actively working on defining the extent and quantities of it. I have been adding to my position this week $zinc $lead $gold $silver $gypsum $emo.v $fwz.v $teck $nexa $glenshow more

introducing a new, very fun, LLM benchmark- the Game-of-Life... Bench! the rules are simple: given an 8x8 grid following Conway's game of life rules, the goal is to create an initial pattern with at most 32 cells that can last the longest number of turns before dying/repeating. some results to highlight (with caveats detailed below): - gpt 5.1 lasts the longest with a 106 step run - claude models are really bad at this! they refuse to reason about this task and score < 25 points - deepseek r1 is the best open model with 102 steps. why? because i wanted to create a benchmark that has (i think) no practicality, but is still fun to look at, cheap, and still measures something interesting. i also am a big fan of the game of life. its absurdly simple rules leading to intractability is extremely cool to me. also, i saw a lot of work with LLMs trying to "predict" the next state in Conway's game of life, I think game-of-life bench is more fun because it's pretty open ended and only asks the LLM for the initial state. I also think this could be an RL env? but idk why you would ever train on this task haha i don't think this is a "serious" benchmark because it doesnt measure anything practical, but i still think it's a hard benchmark exactly because you can't predict what happens with your initial state many turns into the future; this is why i was initially expecting all LLMs to be bad at it, but turns out, some are clearly better than the others (the ordering may surprise you!) reminder: this is still a work-in-progress; (1) i am gpu-poor so could only do 10 runs for each model, even though total running cost is relatively low. maybe with some more credits i can run more seeds for each model. (2) i handpicked models which i think are at the frontier right now, plus some others that were on my mind. so, if you'd like to see a model on here, let me know. (3) i currently only do an 8x8 grid because i thought that by itself would be pretty hard for current LLMs, but of course we can increase grid sizes! (4) the coolest thing is, i dont think we can calculate the max possible number of states (yay undecidability!) you can go without repeating, so this is essentially a no-ceiling task, which is pretty cool! again, i did this mostly out of a desire to make LLMs do something fun. if this keeps me entertained for a few more days, i'd likely release a blog post on it. if it keeps me entertained for a week (and someone sponsors me), i'll put more work into it :P lastly, this is fully open sourced, so feel free to run this on your own!show more

INTRODUCING OCBTW: THE FIRST PANDA-BACKED ASSET A FIRST-OF-ITS-KIND 2-WAY... NFT NFT SWAP We just deployed a smart contract on Bitcoin that allows you to mint "OCBTW", a limited edition generative art collection by tclow.sats, solely by swapping an Alkane Pandas ⬢ OCBTW has a supply of 200 unique pieces and is inspired by Oyl | Building Alkanes's signature hexagon logo that defines all native Alkane assets (e.g., $DIESEL). It is 100% on-chain generative art using three.js code via HTML ⬢ OCBTW can only be minted by swapping a Panda using the 2:70104 smart contract. After 200 swaps have been completed, on a first-come-first-served basis, the mint will conclude. However, OCBTW can be swapped back to a Panda at any time, freeing up supply for another Panda holder to collect a OCBTW piece from the contract. ⬢ OCBTW is thus backed entirely by Pandas, meaning each piece will never be worth less than a Panda. This is the first time on Bitcoin that an NFT is minted with another NFT. This may be the first time this has ever been done on any chain. ⬢ OCBTW also effectively locks up up to 200 Pandas for perpetuity. Just like AP-69, this swap contract has no withdrawal function. The only way to get Pandas out of the contract is to swap your OCBTW back to a Panda. This operates on a last-in, first-out (LIFO) basis. ⬢ OCBTW can be minted by calling the 2:70104 contract using opcode 42 and including a Panda in the transaction, either by using or You can only mint 1 OCBTW per transaction. Rarity of Pandas has no effect on what OCBTW mint you will receive. ⬢ OCBTW can be viewed natively in browser on iDclub 💥 Building Alkanes at the following link: Please be patient after minting for their indexer to update after blocks clear. Please also note that the Ordiscan Alkanes indexer is currently down. ⬢ OCBTW is purely art. Art on Bitcoin. Forever.show more

Ethereum Strategic Reserve & Launchpool Announcement 🚀 💰 Ethereum... Strategic Reserve - 410 ETH StarHeroes DAO is thrilled to announce the creation of the Strategic ETH Reserve, seeded with 410 ETH (~$1.5M) from the DAO’s decentralized Treasury. Inspired by visionary moves from industry leaders like SharpLink Gaming and GameStop, this reserve positions StarHeroes as a pioneer in accumulating ETH while bolstering decentralized infrastructure. The reserve will support: ✅ Ethereum staking ✅ Restaking ✅ On-chain storage More than just accumulating ETH, this initiative is designed to enhance the utility and long-term value of the $STAR token, directly rewarding decentralized holders and strengthening the ecosystem. 🫂 Community Launchpool - 50 ETH Distribution From thrilling in-game seasons to competitive esports leagues and electrifying IRL grand finals, StarHeroes has always been about delivering epic rewards to our community. Now, we’re taking it to the next level with the Season 1 Launchpool, distributing 50 ETH (worth ~$185,000) to $STAR stakers! This is the first time that the entity holding the Strategic Reserve is directly transferring value to the Token holders! 🚀 50 ETH locked in a galactic vault, ready to be unlocked by staking $STAR! 📅 Launch Date: Tuesday, July 29, 2025, at 15:00 UTC—one day before Ethereum’s 10th birthday! Stay tuned for more details and get ready to stake your $STAR for out-of-this-world rewards!show more

Welcome to CyberNetwork, where the future of gaming meets... the power of blockchain technology. CyberNetwork is revolutionising the gaming industry with a decentralised platform that offers unparalleled opportunities for gamers, developers, and investors. By leveraging blockchain technology, we ensure transparency, security, and fairness, transforming traditional gaming models into more immersive and rewarding experiences. CyberNetwork is built on SovereignChain technology from Multiversᕽ, featuring: ⚡️ 100K Transactions Per Second (TPS): Ensuring fast and seamless interactions. ⚡️ 1-Second Block Timing: Providing near-instantaneous transaction confirmations. ⚡️ 2-Second Finality: Guaranteeing rapid settlement and reduced waiting times. This robust infrastructure supports our mission to create a more equitable, transparent, and rewarding gaming ecosystem. Leveraging #MultiversX's Interoperability layer, CyberNetwork is able to seamlessly interact with major blockchains like #Bitcoin, #Ethereum, and #Solana. CyberNetwork Features: 1. Gaming Launchpad 🚀: Empowering game developers to launch and scale their projects. 2. Game Store 🏪: A comprehensive game distribution platform for discovering and purchasing games. 3. Digital Asset Marketplace 🎨: Facilitating the creation, buying, and selling of in-game assets. 4. DEX 💱: Enabling decentralized exchanges of gaming tokens and assets. 5. Gamer's Passport 🎮: A unified identity and achievement system across games. 6. Metaverse Services 🌌: A service for game developers to create high-density virtual worlds. CyberNetwork is set to transform how we experience and engage with games. In the upcoming days, we’ll be unveiling each of these exciting features one by one. Stay tuned for a journey of discovery! For more info: Telegram Discord #Web3Gaming #BlockchainGamingshow more

PERP DEXs & This Week Let’s take a quick... look at what’s happening in perp DEXs this week: RISEx plans to move to mainnet after a reward-based testnet competition on its own chain. The project will launch with limited access codes, so I recommend trying to get one. A major new player is about to enter the perp DEX market! Reya’s trading competition ends tomorrow. The team has also launched a new ambassador program. We’re getting closer to TGE on Reya. Miracle has integrated Nado into the system. After HL and Extended, this is the 3rd perp DEX added. Hibachi launched its first vault. GAV is currently yielding 35%. They surpassed 200k TVL in a short time. One of the most underrated perp DEXs. Pacifica is offering a 50% fee discount on RWA products until March 31. It’s actually quite easy to earn points from OI. On Paradex, fiddy.dime - priv/acc 🦡’s “price shame” is getting worse. The project sits at only a $5M mcap and $26M FDV. Lighter’s $LIT is now below $1. Their “genius” CEO still doesn’t understand that a project is nothing without its community. After TGE, they tried to dump on loyal holders who supported the project for over a year and forced them out with price pressure. Many OGs left. Now he can buy his trash coin at whatever price he wants. Good for him. edgeX🦭 will have its TGE on March 31. Hopefully it will be a good one and set a positive example for the perp DEX market. Based will have its TGE on March 30. The allocation to the community is extremely low, and despite criticism, they didn’t revise it. Expectations dropped from 300–500M FDV to 50–100M FDV on Polymarket. The team is ignoring everything. On TGE day, they’ll likely lose their last blind supporters. They didn’t even open the blind boxes they distributed. That’s how confused they are. StandX — I think the overly positive sentiment and “TGE is coming soon” expectations are unrealistic. There are only 4 pairs. Trading is exhausting, expensive, and difficult. The system needs improvement. They’re not ready, and progress on the perp side is oddly slow. A fast TGE is not possible under current conditions. There are fewer serious traders than you think. If someone inactive for weeks in the top 1000 only drops 10 ranks, it means participation is low. A warning to AG.stand: you are falling behind competitors, and the world is not just Asia. Grvt delaying TGE was a tough decision and of course had consequences. I assume the team calculated this. But if they delay again, they’ll likely face backlash from the community. Variational — Looking at OI and volume, it’s clearly one of the surviving perps. It’s easier to earn points now compared to a few months ago. BULK — I’m not farming it, but White Rabbit is. The testnet is now public. According to them, it’s a fast and well-built platform. Nado has a strong following. I sold my NFT. If they allocate only ~8% to the community, it could cause serious backlash. Hotstuff — I’m not in, but White Rabbit is farming it. They launched a new builder program. Last week, 2,359 users received points. Still relatively low participation. There are constant competitions — including trips to Thailand and other rewards. But pairs are still limited. HelloTrade is still stuck in whitelist phase. Extended is one of the projects managing its process well. If they execute a clean TGE, it could set a strong example. Architect — I haven’t seen anyone actually using it. Feels like Alice in Wonderland. ‼️ 01.xyz — We contacted them about an issue a month ago, and they still haven’t resolved it or even responded that they can’t. With this level of communication, it’s a huge red flag. If I were the CEO, I’d fire the entire communication team. Backpack 🎒 🚮 Hyperliquid 👑show more

Yesterday at 3 AM Claude Code called me I... woke up, picked up the phone, and on the screen was a message: "Wallet entered BTC Up at 11 cents. Open Polymarket?" I said yes and went back to sleep Claude Code unlocked my 2nd phone on its own, opened Polymarket, found the right market, entered the amount, and hit Buy. I could see all of it in real time through the web interface on my laptop. Screenshots from the phone updating every second. By morning the position closed in profit Let me tell you how I got here A week ago I asked Claude Code to write a script that pulls on-chain data from Polymarket and ranks wallets by win rate on 15-minute BTC markets In 20 minutes I had a table with hundreds of addresses, and 1 of them stood apart from the rest. More than 200 trades per day, surgical entry precision, and a profit curve going straight up I fed that address back into Claude Code and asked it to break down the strategy. Turns out the wallet monitors BTC volatility on Binance and Bybit every 100 milliseconds, and when it drops below 0.08% it enters Up and Down simultaneously at 25 to 35 cents A pure straddle: 1 side burns and the other flies to a dollar, giving 3 to 4x per position. Dozens of times a day I wanted to follow it but signals came at any hour, and waking up every 15 minutes for a notification was simply impossible. So I built something else Took an old Android phone and installed an agent running on the Qwen3-VL visual model. It sees what is happening on the screen and mimics human actions through ADB: taps, swipes, text input. Then I connected it to Claude Code as the executor Now the chain works like this: Claude Code monitors the wallet, sees a new position, calls me. And if I say "yes" or just do not pick up within 30 seconds, the agent on the phone opens Polymarket on its own and copies the entry Essentially I built myself an autopilot out of 2 AI systems: 1 thinks and the other presses buttons. I just sleep and occasionally pick up the phone → Here is the wallet the whole thing is tracking: For those who do not want to build a setup like this there is a Telegram bot that handles the 1st part: tracks this wallet and sends a signal on every new entry: AI calls me at 3 AM to ask permission to spend my money A year ago this would have sounded like schizophrenia. Now it is just Tuesdayshow more

Times Square - New York City 🇺🇸 The origin... of this place was known as Long Acre Square, but it was renamed Times Square after being located in 1904, at the intersection of 42nd Street, the offices of the New York Times. In 1913, the newspaper’s offices moved to 229 43rd Street, and the building was renamed 1 Times Square in 1961. History of Times Square, at the beginning of the 20th century, Times Square became one of the most important places in New York City, with the construction of theaters, restaurants and luxury and first class hotels. The site began to change in the 30’s where it became a dangerous place with prostitution, crime and drugs, this lasted for more than half a century. History of Times Square During the 1980s, Mayors Ed Koch and David Dinkins began a redevelopment project for Midtown Manhattan that included Times Square. In the 1990s, Mayor Rudolph Giuliani tackled the project to clean up the area with the closure of pornographic theaters, sex shops and the purchase of the nine historic theaters on 42nd Street by the State of New York. Times Square is located between 7th Avenue and Broadway, and 42nd and 47th streets, it is currently the heart of the city in general and of the New York theater district in particular. The entire area is illuminated day and night by large illuminated signs and huge television screens, this has made it famous throughout the world, as well as the descent of the Crystal ball every December 31 from the top of 1 Times Square announcing the new Year. Up to two million people flocked to Times Square to ring in the year 2000 with the lowering of Waterford’s Millennial Crystal Ball, a huge Swarovski crystal ball made specifically for this event. The north end of the square, between 46th and 47th Streets, is known as Duffy Square in honor of Father Francis P. Duffy, chaplain of the New York Regiment of Infantry during World War II. Half-price same-day Broadway show tickets are available at the TKTS Box Office, located in the center of Duffy Square. The ticket offices, renovated in 2008, are covered on its roof with wide steps in the form of a staircase that allow visitors to sit or enjoy a panoramic view. On May 25, 2009, the city council temporarily closed the section of Broadway that runs through Times Square to pedestrian traffic. The experiment was a great success and it has become a permanent pedestrian plaza where the place is more enjoyable with new metal benches, chairs and tables. Times Square has its own police station, its street cleaning company, and its tourist office that has all the information about the city, and it is also a small museum dedicated to this important square. Times Square wouldn’t be recognizable without its amazing illuminated signs. An image that is admired by two million people who pass through this square every day and many millions more in the world who enjoy it through photographs, paintings, cinema and television. Times Square began to fill with large advertisements in 1916, when the City Council encouraged its installation in this part of the city. Since then there are many who have passed through here, some true works of art. In addition, they played an important role in the remodeling of the square during its most decadent period: as a measure of pressure on the city’s rulers, one night their neighbors turned off all the signs for 15 minutes. The measure had excellent results and the negotiation was carried out. Today, a municipal ordinance obliges the buildings in this area to give up their facade for the installation of such advertisements. 🎥 : Times Square in 1932 (Colorized) 🎥© lostintimevid (IG) #archaeohistoriesshow more

I just sold my startup Talknotes for $200,000 on... acquire.com 💸🤯🤩💰🥳🎉 I launched it last August when I was looking for an idea I could grow with paid ads, and made a MVP in one week. I took it from $0 to $7500 MRR in just 11 months. 👉 Here is how I grew it from zero: 💡 Idea: I got the idea when I tried to write a tweet using Google Doc's transcription tool, but it was terrible. And I was pretty sure I wasn't the one too lazy to type. So I made my own solution, and Talknotes was created. The audience is pretty broad so it was a perfect fit for Meta ads However… ✅ Validation: My rule is to only reinvest what the project generates, so, no ads until I make enough cashflow ❌ Listing on startup directories + a few Twitter sales generated $700 after 10 days. Yes, it's not much, but more than enough to show there is interest in the product and tell me to keep working on it 🤩 I started adding the features users requested, but the launch effect started to wear off and daily revenues quickly went to $0 after a few weeks 🫥 I got depressed and almost gave up on the app... 😔 But luckily, my friends and Dan Kulkov pushed me to continue And I'm glad they did because In October, I launched on Product Hunt 😸 and it blew up 🤯 It got Product of the Day and reached $1500 MRR thanks to the media coverage 🚀🚀 Until then, everything was done using vanilla JS/CSS/HTML + Node for back end. It's simple and easy, but I saw the limitations, so I remade the app using Nuxt to make it easier in the future 🏗️ (thanks to @blackevilgoblin and Piotr Jura for the content/courses! Tim Bennetto as well for the basics!) After that, I took a break and then launched ads on Facebook. The strategy is simple: Catch people's attention, and show them how the app can help them improve their life. No need to over-complicate 🙅‍♂️ Making good creatives is 80% of the job when doing ads on Facebook, most of the technical stuff is done by AI now. Thanks to the boost in traffic, I implemented a feedback loop: 1) Get new users 👥 2) Learn to know them with the onboarding form 💬 3) Make more ads based on the data you get from onboarding 📝 And it completely blew up. MRR doubled in ~2 months However... In May, I had a bad burnout 🥵😩 Multiple bugs slipped into the app, and I had to spend 2 days fixing everything in an emergency while revenues plummeted. This completely fucked me up mentally and had a hard time working on the app after that ( 💀💀 So I decided to list it on acquire.com and made a Twitter post ( I listed it for $200,000, a pretty low price considering the revenues and fast growth. I could have gotten $300,000 if I accepted payment over time, but $200,000 today is better than $300,000 tomorrow for me. 🚨 The process went smoothly until we tried to use Escrow, which almost fucked up the whole deal. (details: I got extremely lucky because the buyer really wanted to buy the app, but this could have ended the deal. We had to wait over a week to get the money back from them, even tho they said they already refunded it. But luckily, after threatening them, they sent it back the next day 🙃 The buyer finally got the money back, I transferred every asset to him, and he sent me the wire. With the profit made from the app + the sale, and other projects, I'm 30% away from being a millionaire 🤯 With this amount, I can pretty much retire in Asia if I want to. But that's just the beginning, I’m going to launch new projects soon! 🚀 But before that, I need to take a real vacation and detox. My brain is completely fucked up by those last 2 months. I gained weight, and got brain rot from scrolling all day waiting for the acquisition to move forward 💀💀 Surprisingly, doing absolutely nothing is 10x more exhausting than working 15h per day 🥱 Now, all this might sound like an overnight success. It is not ‼️ This is the result of 7 years of failure and working like a madman. I launched over 40 projects in those 7 years, and most of them failed. But a few took off, and that’s all I needed All those weeks working 15h/day without weekends and vacation feels soul-sucking when you don’t see the end, but this is what took me there You only need to win once to snowball everything. Work hard, focus, fail a lot and keep shipping fast. 🚀🚀 Thanks to you for reading until here, and thanks to everyone who supported me 🤞show more

Monthly WINR Protocol Development Update: To begin with, the... WINR Protocol has distributed $900,000 to token holders, generated more than $500,000 in pure profit for liquidity providers on WLP, acquired more than 5,000 users, and has almost 9% of the supply burned. In the upcoming months, the WINR Protocol, which has been in production for years, will introduce a range of new products and deployments. These developments will represent the practical and technical evolution to V2 of the protocol. Here are the latest updates and further details as they progress: Progress on WINR Bonanza, Casino Hold'em, and Blackjack is nearing completion. These games are in the final stages of testing. Additional games, including a new type of crash game, have been finalized and are set to debut with the JustBet v2 launch. Take a look at the gameplay videos for a preview. These games achieve the long-term goal of providing a full-suite WINR Game Engine SDK, which can be used to build games with complex logic on-chain with modular smart contract infrastructure. For example, any grid slot game that dominates the iGaming industry could easily be developed on-chain using the WINR Bonanza SDK. - Permissionless Frontend Operator SDK Dashboard Release: March is poised to be a milestone month with the launch of the Frontend Operator SDK dashboards. These dashboards will enable any WINR Labs game to be seamlessly deployed on frontends, marking a significant advancement in protocol accessibility and integration for future games developed by independent iGaming developers. The frontend operator can deploy any game they choose through a few simple steps while utilizing 10,000 WINR per game via the WINR Game Factory smart contract. Each operator is assigned a unique smart contract address(es) for every game they deploy, allowing their revenue to be tracked independently. The deployment process includes instructions on integrating the game as a package into the operator's frontend. In subsequent phases, the games will transition through WINR Chain, streamlining user onboarding steps like wallet connection and token bridging to Arbitrum. This abstraction will make it easy for any web2/web3 platform on any chain to seamlessly integrate WINR-based games with just a few clicks. The new budget system changes how the revenue is calculated on the protocol and will see daylight with frontend operator, Solana, and Fantom deployments. This model was first tested with a lightweight version on and gave a lot of actionable feedback. Shifting from the existing bribe model, which in practice distributes almost half of the edge of the games in volume to WINR holders, the brand-new budget system checks the profitability of WLP. It distributes a larger part of the profit to game providers, frontend operators, and, most importantly, WINR holders. Any time a game's budget is in profit, a part of every loss is distributed to stakeholders. Here is an example: 1. The WINR Bonanza Game has a 10,000 budget for a frontend operator. 2. Let's assume the bet amount was $50, and Bonanza paid back $10 on that spin. That leaves $40 of pure profit. 3. This is distributed amongst 50% to WLP, 20% to frontend operators, 20% to WINR holders, and 10% to game providers. 4. To achieve this, V2 of WINR Liquidity Engine (WLP) will have a buffer for purchasing and selling, working in epochs to determine the above distribution and math. - Solana Deployment and Expansions: Solana audits are in their final phase, and frontend tests are ongoing. Solana launch will be alongside the V2 launch of @JustBetOfficial, with a chain switch available on the top bar. The VRF system WINR developed already is seeing requests from builders around the Solana ecosystem, and this will over time add an extra layer of income to the protocol. Solana's bankroll, at first, will be a lighter version of WLP but will inherit the above-mentioned budgeting system to generate income immediately upon launch. - Fantom Deployment and Expansions: Fantom's upcoming Sonic upgrade, with its 200ms finality and fast block production is a perfect chain for WINR Protocol to expand. Through the partnership with WINR Protocol will tap into a brand new user base on Fantom. The bankroll on Fantom will consist of FTM and stables. The launch is planned for late March or early April. This expansion aims to open up new markets and collaborations with fresh teams. which operates independently, will integrate a broad spectrum of WINR technologies, including WINR Account Abstraction, WINR VRF, and WINR games, marking a pivotal step in WINR Protocol’s journey of horizontal expansion. - XAI VRF Deployment Progress Update: The WINR Account Abstraction Wallet and WINR Verifiable Random Function (VRF) deployment on the XAI 🎮⛓️ is well underway, with the majority of the work completed. This step forward showcases WINR's expansion beyond gambling and trading, highlighting the protocol's adaptability and commitment to broadening decentralized services. The process to automize and permissionlessly let game developers start using WINR VRF by paying (and burning) fees in WINR will launch alongside WINR VRF deployment on XAI and expand to further chains to help DApp builders with the tooling they need. This process will work very similarly to frontend operators, where DApp builders will be able to easily deploy their VRF contract, pay the WINR fees, and enjoy the fastest random number generation transaction, as showcased on @JustBetOfficial for some time. - JustBet v2 Development Update: Set to launch in March, the last testing phase with long-time community members of JustBet V2 is underway. Boasting a completely refreshed look, JustBet v2 aims to captivate more users with its modern features, a significant upgrade from the classic JustBet designed in 2019. Expect dynamic animations and a user experience that rivals traditional web2 casinos, setting a new standard for online gambling platforms and serving once again as proof of concept for all the new WINR features and infrastructure set to launch over the coming months. As always, WINR Labs simultaneously develops protocol infrastructure and platform to best address the needs of one and only goal: horizontal expansion. Easter eggs: Upcoming Gitbook update with all the technical information of WINR V2. CEX listing. Detailed product pages on WINR web. Pyth competitions. WIP-4 and WIP-5 are ready to deploy. And an 🪂show more

What if you kept asking an LLM to "make... it better"? In some recent work at FAIR, we investigate how we can efficiently use RL to fine-tune LLMs to iteratively self-improve on their previous solutions at inference-time. Training for iterated self-improvement can be costly. The naive approach to training for K self-improvement steps leads to K times the number of rollout steps per episode. We introduce Exploratory Iteration (ExIt), an RL-based automatic curriculum method that bootstraps diverse training distributions of self-improvement tasks by upcycling the LLM's own responses at previous turns as the starting points for both self-improvement and *self-divergence.* In order to decide what task to train on next, the curriculum prioritizes sampling of partial turn histories that led to higher return variance in its GRPO group (a learnability score that comes for free). This automatic curriculum over the bootstrapped task space teaches the model how to perform iterated self-improvement while only ever training the model on single-step self-improvement tasks. We look at ExIt's impact in both single-turn (contest math problems) and multi-turn (BFCLv3 multi-turn tasks), as well as MLE-bench, where the LLM is run in a search scaffold to produce solutions to real Kaggle competitions. Across these eval settings, we find ExIt produces models with greater capacity for inference-time self-improvement compared to GRPO. Notably, ExIt models can self-improve on test tasks for many more steps than the typical solution depth encountered during training, including a 22% improvement in MLE-bench performance compared to GRPO.show more

BEIJING JUST BECAME THE MOST DANGEROUS ROOM ON EARTH.... HERE'S WHAT COMES NEXT, STEP BY STEP: Xi just hosted Trump. Then, before the ceremonial trappings were even cleared, Putin walked through the same door. That sequence is not coincidence. It's a message. Here's what comes next, step by step: 🔴 Step 1 — Xi positions China as the indispensable power Trump came first. Putin came second. Beijing received both. That's not diplomacy. That's leverage. Xi is telling Washington and Moscow simultaneously: you need me more than you need each other. 🔴 Step 2 — Putin and Xi sign 40+ documents on energy, trade, and security China is already Russia's largest oil and gas buyer — Europe's pipeline imports collapsed to their lowest levels since the mid-1970s after sanctions. These deals deepen that dependence and lock in the economic architecture of the anti-Western bloc. 🔴 Step 3 — They publish the multipolar declaration The joint statement on "establishing a multipolar world" and a "new type of international relations" is the diplomatic language for one thing: the formal announcement that the U.S.-led order is being replaced. In writing. Signed by both. 🔴 Step 4 — Trump's Beijing visit produces nothing to counter it No major breakthrough on trade. No agreement on Iran. No movement on Ukraine. Trump left with photo ops. Xi immediately invited the man fighting the war Trump couldn't stop. 🔴 Step 5 — The 25-year treaty anniversary becomes a relaunch moment Putin's visit also marks the 25th anniversary of the 2001 Treaty of Good-Neighborliness. They're not just renewing it. They're upgrading it — with energy corridors, military coordination, and a shared ideological framework for the next 25 years. 🔴 Step 6 — Washington's traditional alliance architecture cracks further The back-to-back visits signal that Beijing no longer needs to choose sides. It hosts both. It extracts from both. And it shapes the outcome of both conflicts — Ukraine and Iran — without firing a single shot. 🔴 Step 7 — The real question is what Xi told each of them about the other Xi reportedly told Trump that Putin might regret invading Ukraine. Days later, he welcomed Putin with 40 documents and a joint declaration. One of those statements was for the cameras. The question is which one. The question is not whether this shifts the balance of power. The question is how fast. Are you positioned for what comes next? Follow and turn notifications on.show more

WINDOWS // NYC is an interwoven city that we... will all build out together. There are three collections that all interconnect, windows, buildings, and the city itself. Here is exactly how this will unfold with Transient Labs on Tuesday, June 16th. 🏢🏢🏢🏢🏢 BUILDINGS are generative grids of windows, assembled through an onchain seed derived from Ethereum block data and each work’s token ID, creating distinct sizes and layouts. My collectors, friends and family, and limited partner communities can mint these BUILDINGS before public at 0.015Ξ This presale window is FCFS will last two hours (12 NOON EST to 2 PM EST) All remaining supply will go to public at 0.018Ξ after the presale ends. Supply is 1/1/4000 total buildings, each grid is completely unique, and some have very special qualities :) 🪟🪟🪟🪟🪟 Individual windows are 1/1/1000, and each mint is one single window. This is gated at first to my existing collectors, max mint 50 windows per wallet. Priced at 0.03Ξ, blind mint with instant reveal- this section is packed with gems- cats, plants, neon signs, and a bit of New York City embedded in every frame. Its these single windows that are the foundational layer of this project, since all grids are comprised of these windows. All my existing collectors get these multipliers on mint day, final snapshot is MONDAY the 15th at high noon, EST. For example: Drips // 2x DRIVE Lap 1 // 3x DRIVE Lap 2 // 2x 1/1 // 5x Editions // 1x GRAILS from DRIVE and DRIP DROP have an additional allocation on top. To ask about specific collections of mine, drop a comment here. 🌃🌃🌃🌃🌃 Additionally, the major work of this collection is the CITY, a 1/1 interconnected dynamic galaxy of light composed of all the layers below. This also comes with a large, museum quality physical lightbox of all 1,000 windows illuminated from within. This work will be placed by private sale. The CITY contract will also hold 20 NEIGHBORHOODS, our second largest arrangements of WINDOWS, and can be customized in collaboration with the artist or accepted as random grids. NEIGHBORHOODS will have 15 tokens at 1Ξ for private sale, with 5 held back for top holders of BUILDINGS. I am so excited to build this city with all of you. I've never worked so hard on anything in my life, and I am so grateful to all of you for the support so far. Let's do this thing!!!!show more

Let's talk about agentic product design. Every company has... its own design process. What has always worked for me is spending long studio hours with our product team, dissecting things into pieces and putting them back together. In those sessions we look at value, usability, simplicity, aesthetics, behavior, storytelling, generics, and emotional mapping. I've been crafting products this way for as long as I can remember. Product work at Lemonade isn't for the faint of heart. This obsession over every detail is hard work, but I believe it yields better results and builds stronger talent. One of the things I love about our design and product team is how this process became a second nature to them. Feedback is fast, professional, and tension free. But in our latest session, something was different. One of our designers used Figma and Cursor to build a mockup that was so advanced, it was almost ready to be shipped. It was an incredible glimpse into a world where a single designer working on top of modern low code infrastructure will be able to launch production grade experiences for products with millions of customers, and with LoCo, I expect this to become a reality at Lemonade in just a few quarters. But there's a problem to watch out for. An interesting phenomenon I've noticed over the years is that the higher the fidelity of the work being reviewed, the more defensive people become. When someone shows up with something polished, they tend to resist feedback. They've already fallen in love with what they built, and it's hard for them to accept rejection. Radical candor feedback works best at an early stage of the project, before people get attached and feel the need to defend their work. This session was no exception. Because the work was so advanced, the review became binary, and its maker became defensive. Happily, we all caught ourselves in time to acknowledge this new dynamic and started figuring out how to go back to obsessing about every corner radius, shade of white, and word. When reviewing agentically coded designs, we'll try having our designers bring in more than one option, as well as the open Cursor project so we can make changes in real time if needed. We'll see how it goes, and if this is of interest, I'll update what we learn.show more