Charles Miglietti

AI Innovation Taking Center Stage The recent news of /dev/agents raising a staggering $56M seed round at a $500M valuation is a testament to the transformative potential of AI-driven developer productivity tools. This massive funding round highlights: -The growing demand for AI agents that streamline workflows, enhance efficiency, and supercharge development teams. -How investors are betting big on technologies that redefine traditional software development processes. As someone passionate about leveraging AI for impact, this news reinforces the importance of continuous innovation. Tools like these aren’t just about productivity—they’re shaping how the world builds technology. Kudos to the /dev/agents team for this milestone! What are your thoughts on the future of AI in development? #AI #StartupFunding #Innovation #TechTrends https://lnkd.in/gXnd5DaB

4

That feeling when you realize that the key to testing gen AI at scale is... gen AI! mabl now allows you to include generative AI assertions in your automated tests. 🔥 🔥 🔥 https://lnkd.in/eBG2R_ky

60

4 Comments

We're thrilled to announce that Epsilla (YC S23) now supports Claude 3.5, the latest LLM by Anthropic that’s set to outperform GPT-4o! 🚀🤖 How does Claude 3.5 really perform against GPT-4o? We've recorded a fun video where Claude 3.5 and GPT-4o go head-to-head, tackling the question: Compare GraphRAG with traditional RAG, grounded by 361 research papers on Retrieval-Augmented Generation published since 2024. Curious about which model reigns supreme? Watch the video and join the conversation! For more in-depth insights, check out the detailed comparisons here: Claude 3.5 Insights: https://lnkd.in/gYQBKqrz GPT-4o Insights: https://lnkd.in/gyq38cwA Which one do you think is better? Please share your thoughts in the comments. #Epsilla #Claude35 #GPT4o #AI #ML #Anthropic #LLM

51

Llama goes multimodal! Impressive results from Llama 3.2. Adapters are used for adding in the vision capabilities while keeping the text weights completely frozen to preserve text performance. Also interesting top see CLIP style data of (image text) pairs featuring in aligning the text and vision. Different levels of quality are used for the different stages. Unclear what the vision encoder was but Meta has had some previously good CLIP models with MetaCLIP. Full link in the comments.

47

3 Comments

Today, Vincent Weisser and I are excited to share what we’ve been working on. Prime Intellect is building the infrastructure for decentralized AI development at scale. We aggregate global compute resources and enable researchers to collaboratively train state-of-the-art models through distributed training across clusters. https://lnkd.in/d39i9ux8 We’ve raised a $5.5m seed round, co-led by Distributed Global and CoinFund, with participation from Compound, Collab+Currency, Protocol Labs, and an incredible group of angel investors incl. Clem Delangue 🤗 (HuggingFace Founder), Dylan Patel (SemiAnalysis) and many others. After reading the Scaling Hypothesis in 2020, I shifted my entire focus to LLMs. I joined Aleph Alpha over two years ago, helping build a new, distributed training framework with state-of-the-art efficiency. Following endless discussions on the latest AI research, traveling together to ICML, NeurIPS and more, Vincent Weisser and I decided four months ago to go all in on Prime Intellect to build our vision of an open and decentralized future of AI. Seeing that even the largest organizations begin to train their models across multiple clusters and scale up to hundreds of thousands of interconnected GPUs, it became obvious to us that the open-source community must overcome the limitations of traditional compute infrastructure and find ways to train on thousands of distributed clusters across the globe. Learn more about state-of-the-art decentralized training techniques and our development plans on our blog: https://lnkd.in/dCRjQwfF Our masterplan is as follows: 1. Aggregate global compute (live in beta) 2. Enable globally distributed training across clusters 3. Collaboratively train open AI models in high-impact domains 4. Create a decentralized protocol for collective AI model ownership Our platform aggregating the most cost-effective GPU providers is live in private beta! Sign up for early access at https://primeintellect.ai/ Next, we'll build an open, decentralized training framework to enable collaborative model development across globally distributed GPUs. We’ve started assembling a world-class team across AI foundation model training, novel transformer architectures and distributed systems to execute on this mission. If you’re excited about what we are building, we are actively hiring for founding team members! https://lnkd.in/dxH8WPj8 I'm immensely grateful for the hard work of our early founding team Sami Jaghouar, Kemal Erdem, Johannes Weniger, Mario Sieg and contributors Leonardo (Openprest), Max H., @goth600 over the last months! Back to building 🫡

66

13 Comments

🚀 Exciting News! With Med, we just applied for the SPC Founder Fellowship in Fall 2024, and we highly recommend that other early-stage startups do the same. The SPC Fellowship offers an incredible opportunity to connect with a vibrant community of founders, gain valuable mentorship, and accelerate your startup’s growth. If you’re an early-stage founder looking to take your venture to the next level, don’t miss out on this chance. Check out their blog in the comments to learn more about the fellowship and how to apply. #startup #entrepreneurship #innovation #fellowship #SPC #earlystage

1

4 Comments

I am coming to Boston! I don't remember the last time I visited Boston, but I can't wait =) 🔉 Calling all founders. Join me next Wednesday (April 24) at this workshop "How to Pitch Your Startup 101". I will be judging at Techcrunch Early Stage 2024 the next day 👉 https://lnkd.in/gYyzvMBU ​There is "no love at first sight" in startup pitching. An investor will not, cannot, and would never invest in you the first time they meet you. ​At this talk, I will cover: - ​What do investors want? - ​What do founders want from your 1st pitch meeting? ​- What should a founder say and do in the first pitch? ​- What should a founder NOT to say or do in the first pitch? ​- What should a founder ask a potential investor? ​- What should a founder do to follow up? Sign up here: https://lnkd.in/gK9Yvmya #venturecapital #funding #pitching #founders #startups #investors #pitchmeeting #fundraising #harvard #MIT #techcrunch #VC #pitch #pitch101 #investing #TC cc Calvin Chin, Pracheer Gupta, Aidan Ryan, Jonathan Kay, Eliran Sapir, Race Capital, Dave Schappell, Brian Skerry, Adam Honig, Luke Stephan

67

13 Comments

5 hours long Lex and Dario Amodei, the co-founder and CEO of Anthropic, a company behind Claude 3 Haiku, Claude 3 Sonnet, and Claude 3 Opus LLMs, speak about how AI is shaping the future and how it could impact the way we live and work. They discuss how far AI can go. Amodei thinks AI can match human intelligence and even surpass it in areas like biology, where complexity overwhelms humans. But in other areas, like speech recognition or conflict resolution, AI might hit limits close to ours. One of the challenges is the fact of running out of quality data and the cost of scaling compute power. Enjoy the entire talk! https://lnkd.in/dRhpafzU

9

2 Comments

The accuracy isn't great (so far), but the responsiveness and efficiency of Moshi are incredible. Try it out here: https://us.moshi.chat/ #AI

8

6 Comments

One more intresting presentation of Paul Tremblot de la Croix "How Generative AI Can Help Startups Succeed" at the #VDS2024. And my note on how AI can drive success: 💬 speakers emphasized that data is the foundation. Generative AI applications are only as good as the data they're built on. The ability to combine real-time data with generative capabilities is a game-changer for startups. 🕵‍♂️ shared the importance of customization. Don’t adopt generative AI just because it's trendy. Tailor it to your business needs, ensuring it’s the right tool for your use case. 👮‍♀‍ discussed data privacy and security concerns, emphasizing GDPR and other regulations. Startups must handle customer data carefully and integrate strong security protocols when deploying generative AI tools. Key points: 1. Generative AI + RAG = Personalization. By using AI with RAG, startups can retrieve relevant real-time information or past interactions, ensuring personalization not just based on predictions but also on actual preferences, history or external knowledge sources. 2. Customization & Use Case Fit. It's crucial to pick the most efficient AI model, considering cost, latency and accuracy. Generative AI isn't one-size-fits-all! 3. From MVP to Production. Building an AI-powered product is just the first step. Measuring, tracking and fine-tuning AI in production is essential for maintaining customer trust👌 I appreciate your meaningful insights shared about scaling AI strategically. Also on keep security tight and focus on solving real customer problems and think about there needs 😇 #AI #Startups #VDS2024 #DataPrivacy #Cybersecurity

4

HITRUST chose Vanta as the first pre-built HITRUST e1 solution, which means Vanta comes "out of the box" with the necessary controls, documents, and policies for an e1 assessment and eliminates the manual “do-it-yourself” approach that other platforms require. Up to 80% of requirements are automated by: *️⃣ Guidance around requirements: 44 new controls and 72 automated tests to ensure comprehensive and continuous compliance. *️⃣ Documents and policy templates: 80 new documents and 10 policy addendums to outline practices around managing sensitive data. *️⃣ Automated evidence collection: With over 300 integrations, Vanta automatically and continuously collects evidence from an organization’s technology stack. Our platform also applies overlapping, implemented controls from other supported frameworks, including up to 50% of SOC 2 and ISO 27001, to eliminate duplicative work across compliance programs. Coming soon, our partnership with HITRUST will expand into an integration with their audit portal, MyCSF, so customer evidence is transfered automatically from Vanta to MyCSF, further streamlining the HITRUST validation process Really excited for this collaboration!

251

14 Comments

🚀 Nuclia delivers the first-ever open-source RAG evaluation model—a crucial tool for the entire RAG community! 🚀 Retrieval Augmented Generation (RAG) methods have revolutionized document search, leveraging the latest advancements in Large Language Models (LLMs) and Deep Learning. However, evaluating these complex systems can be challenging due to the intricate nature of LLM outputs. At Nuclia, we’ve addressed this challenge with REMi, our efficient open-source LLM fine-tuned specifically for RAG evaluation. Alongside REMi, we’ve also developed nuclia-eval, an open-source library designed to simplify the assessment of RAG pipelines using REMi. https://lnkd.in/dBPHgayA

4

It’s encouraging to see Llama 3 has such good performance. I wonder how other large open source model do (if they participate) on the leaderboard. It will be great to see Grok 1.5 also evaluated here.

6

We are thrilled to announce that TurboDocx has successfully closed its Angel funding round, raising $75,000. This significant milestone marks a leap forward in our journey to revolutionize smart and modular document and slide deck creation, accelerated by AI. A heartfelt thank you goes out to our friends, family, early supporters and everyone that believes in our vision. Your support is not just appreciated-- but a strong vote of confidence in our mission to innovate and transform how business document creation is created. What does this mean for TurboDocx? We're reinvesting it into the business: - Focusing on Partner/Client Needs - Understanding and addressing the evolving needs of our users remains our top priority. We're here to build not just a product but a solution that truly makes a difference. - Enhanced Solutioning - We're more committed than ever to developing solutions that set new benchmarks for efficiency, re-usability, and innovation has never been stronger. - Innovation, Partner, and Channel Growth - Our investment team and advisors primarily consists of go-to-market (GTM) and implementation professionals that have already provided invaluable guidance on scaling our overall growth. Looking ahead: As we celebrate this achievement, we're mindful that fundraising is not the end goal but a milestone --a means to an end-- to build out a well oiled machine. We're excited to leverage this opportunity to further our mission, drive innovation, and exceed the expectations of our users and partners. With this said - If you'd like to be part of this exciting journey, please reach out! Stay tuned for updates, join our growing community, and let's reshape the future of document and slide deck creation together. -Nicolas https://lnkd.in/e4CGrUVB

178

40 Comments

Our Kids & Tech Guidebook launched today! If you have a kid in elementary school this book is for you. It lays out 10 practical steps to help you navigate tech choices for your kids. You can download the FREE guidebook at the link below in my comments. Super easy. Check it out!

27

2 Comments

Is this the end of embeddings? Hear from Josh and Tashrish on the pros and cons of using embeddings-based RAG for document data extraction and our experiments with switching to completions-only RAG.

5

OpenAI buying Rockset is massive news in the #realtime data space, but the question is what it actually says? First question : What did Rockset actually solve for? By their own marketing, they power "in-app search and analytics" they compare themselves ClickHouse, Imply(Apache Druid), StarTree (Apache pinot) and Elastic search. Amusingly, Clickhouse and Tinybird have gone aggressive on ads about migrating from it. This places them squarely in the realtime search/analysis space. Considering how much data recall large language models require, the usage makes more sense. My experiences with Rockset have mostly been with a few very impressive demos involving massive ingest and queries of JSON data. It very much felt like what Solr and Elastic search felt like. Second question: Why did OpenAI buy them, and sunset the cloud service? I have a number of theories, but mostly it comes down to this: It would cost OpenAI more than 20% of the purchase price PER YEAR to just use the product as is. I'd guess their spending FAR north of 100m a year on just data services. As for sunsetting the cloud services? They likely cost more to operate than they make. Most startups have a rough time on margins, so maintaining a few tens of million in revenue doesn't make sense when they could re-task the team to power the Open AI rocket ship. Building proper multi tennent data isolation tools is HARD, and Rockset had some of the best low effort ingestion out there. Last question for today: What does this mean for the space at large, and for their customers? Cynically, its a big validation for most of the big investments into the realtime analytics & search space. There are a lot of large players in racing to corner the LLM space and with Tabular (now part of Databricks) and Rockset exiting, This will cause some portfolios to spike. HOWEVER! This does show why open source/core tools could be a more strategic move going forward, and moving off tools like this are a pain. Rockset built a great product, and now its gone. Selfishly, I'd like to see more of this work move into #apachePaimon, and #apacheFlink but I am biased :D. For the recall, enrichment, and training pipelines, having exactly-once stream and batch semantics on a realtime #datalake really adds some reactivity to the user eperience :) references: https://lnkd.in/dY4JHecj #apachedruid #apachepinot #bigdata #LLM #openAI #chatgpt

54

5 Comments