Bui Dinh Ngoc

Shibarium Witnesses 12% Surge in Total Value Locked (TVL) Amidst Ongoing Developments Shibarium, the layer-2 scaling solution for the Shiba Inu ecosystem, has recently seen a significant increase in its Total Value Locked (TVL) from $2.15 million to $2.42 million. This growth, representing a 12% increase, reflects a substantial surge in the platform's usage. Notably, the TVL had been less than $1 million at the beginning of January. The leading decentralized exchange on Shibarium, Woofswap, accounts for $1.68 million of the TVL, with other platforms such as ChewySwap, MARSWAP, DogSwap, and PunkSwap also contributing. The migration of ShibaSwap from Ethereum to Shibarium is anticipated to bring about a considerable escalation in Shibarium's TVL, which currently stands at $28.9 million. Additionally, a recent hard fork upgrade has been implemented to enhance speed, reduce costs, and stabilize gas fees in order to bolster Shibarium's performance within a competitive landscape. # Thank you Elena Silva for your submission!

1

Key Insights After Attending #Devcon2024 in Bangkok: [Part 1] 1/ $16B Cash Buying Pressure from FTX claims in Q4 2024 Mark the 2 remaining dates of “omnibus hearings” scheduled for November 20, and December 12 and check what happened after the first one on October 22. All the way up IMHO, BTC to 180k in Q2'2025. NFA 2/ VC Challenges Many venture capital firms remain heavily illiquid. Raising N+1 funds remains difficult for existing funds due to poor performance. I expect to see plenty of new funds appear in Q2 2025 leveraging BTC price action but they'll be late to the party and we will not see them in the next cycle. 3/ Monad vs. Solana -> I do Believe, MegaEth vs. Solana -> I do not Monad Labs made it clear at Demo Day - "we want to challenge Solana Foundation and we'll grind for highly transactional retail-focused businesses - betting and gaming. Keone Hon and team CULTivate 4/ DeSci is a Category to Watch out CZ met Vitalik at DeSci event. Speculation is rife about how many pitch decks are being drafted in this category. If one gets funded, it could be a banger. 5/ Devcon Holds Tier 1 Conference Status but it's not longer Ethereum only The event boasted an incredible venue, over 700 side events, and a strong presence from major ecosystems like Solana, Sui, Monad, Base, Bera, Ton, and Polygon. 6/ Panel Focus: Tech-centric A clear need exists for panels centered on consumer applications and growth strategies. Considering stepping up as a host of a "Consumer Day". 7/ Infrastructure Projects aim to position themselves as middleware yet struggle to define their target users. 8/ Applications heavily lean towards memes and betting—reminiscent of Ethereum and Tron’s early days. 9/ Top Side Events: Stanford BASSLite for quality engagement, Crypto Pool Party by JustMoney x Twiskers for seamless UX https://lnkd.in/dJu4VEai https://lnkd.in/d-nJ6Giz The highest-quality sessions were hosted by Stamford, where attendees often stayed engaged for up to six hours. I have never seen that many VCs staying in one place for so long 10/ Avoid Mentioning You're an Angel or VC: Keeping your investor status undisclosed leads to better, more genuine conversations. "Wanna-be" founders quickly lose interest, while truly bullish founders drop the pretense and become more authentic. 11/ Word App's Potential: Worldcoin reportedly delivers KYC-verified users with zero marketing spend, a potential game-changer for user acquisition. 12/ Thailand Observations: Notable extremes: high humidity, traffic, stark social disparities, and local cuisine that, while good, feels overhyped (especially by Western attendees). I'd not live there but spending a week or two is worth it.

8

2 Bình luận

Update 🚨 @Certik report for @Saitachaincoin is 100% completed! In just 3 days, they will be publishing the score. Meanwhile, everyone can check the detailed report published by Certik https://lnkd.in/eQsABkzq We are eagerly looking forward to sharing the final score with you all. Stay tuned for updates and thank you for your continued support! #Saitachain #Certik #Blockchain #Crypto #stc #sbc #bugbounty #web3 #blockchain

17

3 Bình luận

NODO Blockchain Expert Meetup Venue: BaoBox, General Mathenge Close Date: April 20, 2024 Saturday at 2 PM sharp found myself seated at Bao Box, located at General Mathenge Close—an ideal spot to hang out with its variety of indoor games and serene atmosphere. However, enjoying the draft beer was not the main agenda; it was a blockchain expert meetup. Being an early bird, I caught up with Faith Nyambura, a @NODO Ambassador. In the early 2000s, the internet evolved into what is known as Web 2.0, where interactive social media apps became the norm. Users could post content, comment, and express likes or dislikes. Unfortunately, big companies reaped all the profits, not acknowledging the multitude of online users who interacted with these platforms. NODO —a Web 3 African news and discovery site—takes a revolutionary and inclusive approach. It aims to disseminate information about Web 3 products while rewarding its users with Gems, which can be cashed out. One can interact with the platform as a user, maker, expert, or content contributor, with every interaction being rewarded. The event started with GRANTON NYANGE, NODO ambassador, moderating. Initially, there was an open forum where the topic of Bitcoin halving and its expected consequences was discussed. Panel Discussion Faith Nyambura moderated the panel discussion. Panelists: Nelly Njeri, Founder of Ledgerlady: Highlighted the community's amazing work in recruiting more women into the blockchain space. ERICK MUNGAI, CTO of @Cause Coin: Discussed the mission of launching the first officeless crypto bank, set to launch later this year in Kenya. Edward Karani of GoChapaa: Talked about empowering Africa to take control of its financial future using blockchain technology. Topics discussed included regulations—with the Blockchain Association of Kenya (BAK) being commended for its efforts in developing the Virtual Asset Service Provider Bill—, and challenges like fewer women in the tech space, intellectual property rights, and access to funding were explored. On the topic of funding, Faith discussed ways to access funds through NODO grants. At the end of the panel discussion, Griffins Oduol from NexusPay App shared his project, which seeks to lower the access barrier to cryptocurrencies by making them accessible to all including users of feature phones. I also had the opportunity to meet Moses Timbwa and Naomi Bett

13

HKU and teams at Tencent research are tackling the evaluation problems of language models with a proposed reasoning benchmark. Plot2Code is a benchmark designed to evaluate the capabilities of Multi-modal Large Language Models (MLLMs) in generating executable code from scientific plots, addressing a previously unexplored area. It comprises a curated dataset of 132 high-quality matplotlib plots, spanning six distinct types, each paired with its source code and a descriptive instruction generated by GPT-4. This setup allows for a comprehensive evaluation of MLLMs' coding capabilities across various input modalities.   The benchmark employs three automatic evaluation metrics: code pass rate, text-match ratio, and GPT-4V overall rating, providing a nuanced assessment of the generated code and the fidelity of the rendered images to the original plots. Unlike traditional binary pass/fail evaluations, these metrics offer a detailed judgment of the output, aligning closely with human evaluations.   Evaluating 14 MLLMs, including proprietary models like GPT-4V and open-source models such as Mini-Gemini, Plot2Code reveals significant challenges, particularly with text-dense plots where most MLLMs heavily rely on textual instructions. This highlights substantial room for improvement in visual coding tasks.   Plot2Code's contributions are multifaceted. It provides a robust benchmark for evaluating MLLMs' visual coding capabilities, setting a new standard for future developments. The benchmark's design, emphasizing a diverse range of plot types and evaluation metrics, ensures a comprehensive assessment that can guide the enhancement of MLLMs. The open availability of the Plot2Code dataset encourages ongoing research and development, offering a valuable resource for the AI community.   The benchmark supports two evaluation settings: Direct Asking, where the model generates code from an image, and Conditional Asking, where the model generates code based on an image and additional textual instructions. These settings allow for a thorough evaluation of MLLMs' performance across different input modalities.   Plot2Code underscores the complexity of visual coding tasks and the need for further advancements in MLLMs. It aims to drive future research in multi-modal reasoning, text-dense image understanding, and complex code generation, paving the way for more intelligent and versatile multi-modal systems.  Arxiv: https://lnkd.in/e2MkTxbD

2

aiXcoder-7B: A Lightweight and Efficient Large Language Model Offering High Accuracy in Code Completion Across Multiple Languages and Benchmarks The research team from aiXcoder and Peking University introduced aiXcoder-7B, designed to be lightweight and highly effective in code completion tasks. With only 7 billion parameters, it achieves remarkable accuracy compared to larger models, making it an ideal solution for real-time coding environments. aiXcoder-7B focuses on balancing size and performance, ensuring that it can be deployed in academia and industry without the typical computational burdens of larger LLMs. The model’s efficiency makes it a standout in a field dominated by much larger alternatives. The research team employed multi-objective training, which includes methods like Next-Token Prediction (NTP), Fill-In-the-Middle (FIM), and the advanced Structured Fill-In-the-Middle (SFIM). SFIM, in particular, allows the model to consider the syntax and structure of code more deeply, enabling it to predict more accurately across a wide range of coding scenarios. This contrasts with other models that often only consider code plain text without understanding its structural nuances. aiXcoder-7B’s ability to predict missing code segments within a function or across files gives it a unique advantage in real-world programming tasks. Read the full article here: https://lnkd.in/gRBm75BK Paper: https://lnkd.in/gkxTKDku GitHub: https://lnkd.in/gGg5dK3R #ai

33

1 Bình luận

LAMBDA: A New Open-Source, Code-Free Multi-Agent Data Analysis System to Bridge the Gap Between Domain Experts and Advanced AI Models A team of researchers from Hong Kong Polytechnic University has introduced LAMBDA, a new open-source and code-free multi-agent data analysis system developed to overcome the lack of effective communication between domain experts and advanced AI models. LAMBDA provides an essential medium that allows smooth interaction between domain knowledge and AI capabilities in data science. This method solves numerous problems like removing coding barriers, integrating human intelligence with AI, and reshaping data science education, promising reliability and portability. Reliability means LAMBDA can address the tasks of data analysis stably and correctly. Portability means it is compatible with various LLMs, allowing it to be enhanced by the latest state-of-the-art models. The proposed method, LAMBDA, a multi-agent data analysis system, contains two agents that work together to solve data analysis tasks using natural language. The process starts with writing code based on user instructions and then executing that code. The two main roles of LAMBDA are the “programmer” and the “inspector.” The programmer writes code according to the user’s instructions and dataset. This code is then run on the host system. If the code encounters any errors during execution, the inspector plays the role of suggesting improvements. The programmer uses these suggestions to fix the code and submit it for re-evaluation. Read our take on this: https://lnkd.in/g6HN4i_W Paper: https://lnkd.in/gjkV9fKv Project: https://lnkd.in/gnq4Zfhr GitHub: https://lnkd.in/gsmyvj9j

52

3 Bình luận

Search, sort, and filter our comprehensive database of 8000+ #VC professionals. #VCFirm Name Website Offices Stages Portfolio Link Markets Firm LinkedIn Twitter VC Name Title LinkedIn Twitter Access Airtable https://lnkd.in/gEjqNNam #venturecapital #angelinvestors #startupgrowth

1

new proxy who this? support of (soon ™️): - rich media type responses - file uploads - file downloads all this and more in our changelog below👇

24

2 Bình luận

Do you agree with me? 🤖 Humanoids and 🥷Superagents will run an: ➡️ AIOS by Kai Mei, Prof. Yongfeng Zhang 📟 https://lnkd.in/gcn9m7mx or/and ➡️ an LLM OS by Anrej Karpathy 📟 https://lnkd.in/geN_agUH Building these OSes and their core components (like tools, schedulers, memory, storage, core) will yield another wave of 🦄 big tech Teracorns 🦄 https://lnkd.in/gnUBQgSe Credits: - Kai Mei, Prof. Yongfeng Zhang - Andrej Karpathy https://karpathy.ai/, Eureka Labs

15

2 Bình luận

𝐇𝐮𝐧𝐲𝐮𝐚𝐧-𝐋𝐚𝐫𝐠𝐞: 𝐀𝐧 𝐎𝐩𝐞𝐧-𝐒𝐨𝐮𝐫𝐜𝐞 𝐌𝐨𝐄 𝐌𝐨𝐝𝐞𝐥 𝐰𝐢𝐭𝐡 52 𝐁𝐢𝐥𝐥𝐢𝐨𝐧 𝐀𝐜𝐭𝐢𝐯𝐚𝐭𝐞𝐝 𝐏𝐚𝐫𝐚𝐦𝐞𝐭𝐞𝐫𝐬 𝐛𝐲 𝐓𝐞𝐧𝐜𝐞𝐧𝐭 Tencent The Hunyuan-Large model (389B total params, 52B activated) outperforms several recent models including LLama3.1-405B and others across multiple benchmarks. Features 256K context length and MOE architecture. 🏆 Outperforms LLama3.1-70B and competes closely with LLama3.1-405B. Achieves 88.4% on MMLU, 92.9% on CommonsenseQA, and 71.4% on HumanEval. 💡 Incorporates KV cache compression, expert-specific learning rate scaling, and a mixed expert routing strategy, resulting in nearly 95% KV cache savings that improve inference efficiency. 📊 Trained on 7 trillion tokens, including 1.5 trillion synthetic tokens. Abs: https://lnkd.in/gbm4juZP HF: https://lnkd.in/gqWRai8w #AI #GenAI #LLM #MLLM #VLM #MOE

2

A new Hyperledger Project for decentralized identity solutions : Hyperledger Identus!! Hyperledger Identus offers a comprehensive toolset aligned with self-sovereign identity standards. This innovative technology streamlines the integration of decentralized identity solutions into existing workflows, providing complete DID and verifiable credential functionality. Hyperledger Identus operates as a layer-2 blockchain solution, utilizing a distributed ledger as a verifiable data registry. Users can now securely manage their digital identities while upholding the integrity of their credentials. Check out Hyperledger Identus on GitHub for more details: https://lnkd.in/dY7D9U8Q. #Hyperledger #DecentralizedIdentity #SSI #Blockchain #DigitalIdentity Getting started : https://lnkd.in/d-RMkqTQ Hyperledger Foundation Hyperledger India Chapter

29

1 Bình luận

A small function of SuperPersonalAgent Of TensorChat. If someone chats with your agent and there is no answer in your database, your agent will automatically send an email asking for your identity.😀Based on Autogen Chi Wang

6

4 Bình luận

🙁 🕹 Why Are Korean Gamers Against Blockchain Games? For the past few years, many Korean gamers have been opposing games integrated with blockchain technology. This trend started after the infamous $LUNA crash in May 2022, which halted the further development of GameFi in the country. What are the reasons behind this attitude? A long-standing distrust of gaming companies like NCSoft, Netmarble, and Nexon, is often accused of holding back technological progress in Korea. The token sale of leading P2E company WeMade was seen as a greedy move to acquire another studio. Legal regulations: When it comes to blockchain marketplaces, legal issues arise. The resemblance between Pay-to-Win and Play-to-Earn models. Fear of economic instability. As a result, the average age of Web3 gamers in Korea has risen. The core user base now consists of men aged 30-40, who are less sceptical about P2E models. It is important to note that the Korean gaming market is the fourth largest in the world. Despite this, the Web3 segment in Korea will develop, but at a much slower pace than in Western countries, where there is no such deep-rooted culture of distrust. Also, Web3 games are not ready to fully adopt the masses of Web2 players and adjust them to their reality. This is evidenced by the recent shift of the game Dimensional into the Web2 space. #BlockchainGaming #GameFi #LUNA #Web3Gaming #Web3Games #BlockchainGaming #CryptoCommunity #GameFiNews #DimensionalGame #P2E #Korea #GamingIndustry #Web3Community #CryptoGaming

8

1 Bình luận

It warms my heart to share this: BION has received investment from TON Ventures But I would like to take a moment to share more. When I was at the Token2049 Singapore event, I saw everyone discussing about the TON blockchain. It was the usual discussion: Telegram has 900 million users, massive distribution network, Notcoin, and how mini-apps are taking over. I knew I made the right decision. And now, with TON backing us, it gets even better: Because with this investment, we are sharpening our focus to reach new heights. To deliver massive value. And to eventually onboard millions of users to web3. This also means: 1. There is a lot MORE to achieve I understand users are the lifeline of every startup. This is why we started small. Bion was first rolled out in India — with 50 merchants — and then slowly expanded to other countries. Now we have over 1000+ merchants and we are live in 30+ countries. Still the beginning. 2. Bion is the first shopping app on the TON blockchain Interestingly, I never knew this. After exploring the Telegram App Center, I realized Bion is the first shopping app live on the TON Foundation. There is no mini-app out there offering cashback for shopping — cashback in $TON, $USDT, and $BTC. 3. We are grateful for the support In my seven years of being an entrepreneur, I have NEVER seen such an incredible community. This web3 community is truly filled with people listening to you. Rooting for you. And pushing you to succeed. Special thanks to Ian Wittkopp Inal Kardan Alena S. Sophia Rusconi for your support, invaluable insights, and relentless encouragement. Also, thanks to Vijay Pravin Maharajan Ramani 'Ram' Ramachandran Harsh Rajat Vivek Kumar Suraj Chawla Santosh Panda Jagdish Pandya ( JP ) Sangeet Muralidhar Rabilal Thapa Divyang Patel being so wholesome throughout this journey. It is reassuring to have such good friends all along. (Trust me, I have named only a few — there are SO many amazing people. And I want to THANK each one of you…) 4. Future Goal: 100 million users As ambitious as it sounds, I want Bion to lead the shopping category in web3. Even though we are an early-stage startup, I truly believe there’s a massive opportunity here. I can go on and on about this but the goal is to onboard *lots and lots* of folks from web2 to web3. 5. This is the beginning! And this beginning was catapulted with Ian Wittkopp’s involvement — believing in us to innovate boldly. He has been instrumental in this entire journey. Such pleasure to have an investor and friend like him. As cliché as it sounds, this is still the beginning and I am excited for what’s next. All in all, over the course of the next couple of weeks, we will be expanding our team, enhancing our tech, and spending more time putting the pieces together. Because, after all, a good startup is built on strong foundation. (PS: We are hiring and growing our team quickly. If you are interested, DM me to join us to shape the future of Web3)

90

42 Bình luận

[Q] What's good about the deep learning research team at Picovoice? [A] We are doing things differently so we don't have to train the same architecture on many H100s while praying that we can magically beat companies doing the same thing with more GPUs. [i.e., we have a real chance of success.] [Q] What's bad about the deep learning research team at Picovoice? [A] We are doing things differently so we don't have to train the same architecture on many H100s while praying that we can magically beat companies doing the same thing with more GPUs. [i.e., we are tackling hard problems with no existing solutions.] BTW we are hiring for three roles within the team https://lnkd.in/ehq_UaJ6 https://lnkd.in/e87CFqJb https://lnkd.in/eh82KAyB Happy Friday, everyone! #hiring #deeplearning #llm #voiceai

27