Adrien Supizet

LLMs are failing in trading. More data + better models = smarter trading, right? Traders are still getting wrecked. The gap isn't intelligence, it's integration.

Glad to have secured an allocation in the future of onchain AI infrastructure credit with $CHIP @USDai_Official Staking $PENDLE is the real alpha

DeFi Analytics Tools Tier List 🧵 S: @dune: Onchain data platform where you query with SQL and build/share custom dashboards across 100+ chains. @DefiLlama: Aggregator for DeFi metrics across chains and protocols. Best for the default macro dashboard to compare protocols & chains metrics with a great UI. @flipsidecrypto: AI + data platform providing labeled onchain data and conversational analytics. Great for ecosystems to understand users, growth, and protocol health without building their own data stack. @surfai: AI copilot for crypto research combining onchain data, social sentiment, and curated reports with in-app execution. Great to filter through the noise and find real alpha. A: @dexscreener: Real-time charts and pool data for DEX pairs across many chains for tracking new pairs, liquidity, PA, and basic onchain trading. @CryptoRank_io: Market aggregator for prices, listings, token sales, unlocks, fundraising, and airdrops. @minara: Financial-focused AI that aggregates portfolios and market data across crypto & stock markets. Top pick for financial reports, research, etc. @tokenterminal: Fundamental data terminal for protocols and chains (revenue, fees, users, ratios). Best to value projects like equities with standardized metrics and ratios. @_dexuai: Analytics platform mixing market, technical, and social data for narrative tracking and determining social sentiments. B: @Tokenomist_ai: All about tokenomics/unlocks analytics, such as allocations, vesting, emissions, etc. @cryptoquant_com: Onchain + derivatives data such as CEX flows, stablecoin metrics, miner/whale behavior. @l2beat: Analytics and research hub for Ethereum L2s, go-to for tracking TVL/TVS, designs, security assumptions, and risk frameworks for L2s. @artemis: Data platform focused on onchain fundamentals and stablecoin metrics with a terminal and good API. C: @coinglass_com: Derivatives and liquidation analytics like OI, funding, heatmaps, etc. @nansen_ai: Onchain analytics with massive wallet labeling. Mostly used to follow smart money wallets, funds, etc. All real insights are in the paid version. D: @glassnode: OG onchain metrics suite for BTC/ETH and macro flows, good quality data, but paywalled, data may be found on other free tools. @dappradar: App ranking and usage tracker, decent for surface-level research, but had some instances of outdated data. @santimentfeed: Combines onchain and social metrics, nice concept, but the data can easily be found in other apps with better UIs and user experiences. What tool did I miss? Hope you enjoyed this post and found some new VALUABLE information. If you wanna support me, I'd appreciate a like, reply, and RT <3

This level of financial detail is how we re-build confidence in liquid tokens Amazing!

Send it! $BTC

The X algorithm should be RL'd with its reward being the Kardashev scale

Without doubt, the 2 most asymmetric opportunities in the world right now are AI and crypto. Do whatever you have to do to master both.

In 3 years, you’ll either: a) Own the AI tools b) Be managed by someone who does c) Get replaced by someone who automated your job Your actions will determine which category you fall into.

@_avichawla Great thread @_avichawla Is there such thing as ethics in model training?

How LLMs train LLMs, clearly explained (with visuals):

The ATH breakout was crystal clear—exchange liquidity levels told the story before the move. That' s the power of data 📊 $BTC #DataEngineering #AllTimeHigh #altseason #crypto

We're excited to announce the successful completion of a cross-chain Delivery versus Payment (DvP) transaction in collaboration with Kinexys by J.P. Morgan (@jpmorgan) and @OndoFinance. prnewswire.com/news-releases/… This milestone marks the first-of-its-kind cross-chain, atomic settlement of a tokenized asset between Kinexys Digital Payments’ permissioned blockchain network and a layer-1 blockchain network. Powered end-to-end by the Chainlink Runtime Environment (CRE), the transaction involved the exchange of Ondo Chain’s Short-Term U.S. Government Treasuries Fund (OUSG) as the asset leg with Kinexys Digital Payments’ network serving as the payment leg. Since inception, the Kinexys platform has exceeded $1.5 trillion in notional value, processing an average of more than $2 billion daily in transaction volume. Payments transactions have grown by 10x year-over-year. 🧵⬇️

1/ Kinexys by J.P. Morgan (@jpmorgan), Ondo Finance, and @chainlink are teaming up to connect bank settlement infrastructure to Ondo Chain. Today, we are excited to announce the debut transaction on Ondo Chain testnet — a cross-chain, atomic DvP settlement of OUSG.

@TheAITimeline Super helpful! Quite interested in ReasonIR: Training Retrievers for Reasoning Tasks @ypwang61 👏🏻

🚨This week's top AI/ML research papers: - DeepSeek-Prover-V2 - The Leaderboard Illusion - Phi-4-reasoning Technical Report - Mem0 - X-Fusion - Softpick - RL for Reasoning in LLMs with One Training Example - ReasonIR - RL for LLM Reasoning Under Memory Constraints - Phi-4-Mini-Reasoning - Beyond the Last Answer - Scaling Laws For Scalable Oversight overview for each + authors' explanations read this in thread mode for the best experience

@pythianism @drjasper_eth Like it or not, #institutions ask themselves: which L1 will keep assets secure uninterruptedly for the next decade? While #Solana gets adopted with great scalability, #Ethereum will keep leading in trust and security.

The only place for large scale TradFi experiments where something like BUIDL can reach $2-3Bn in AUM is ETH L1. That is the only place where the capital markets are deep enough, and the more experiments that take place here, the deeper the capital markets get.

#Stablecoins are booming everywhere. Funny how hard it was to get screen-addicted folks to use internet money!

If you don’t have goals that keep you up at night, You’re probably wasting your potential.

@opencampus_xyz EduFi is a lever for learners and teachers to do more with more, not more with less effort. I like @veritasium's take on this: "Effort is the algorithm".

DeFi changed the world. What do you think EduFi can do?

AI’s turning everyone into part-time experts. Real experts? Multiply your edge with AI and you won’t have to look back. Intuition, experience, or—better than all—your ability to put in effort.