Ashton Hettiarachi

Meta just released LLaMa 3.1, the most powerful open source LLM to date, which outperforms GPT-4 and Claude 3.5 on some parameters. This is the biggest moment so far in the open vs. closed debate in AI. One of the big questions for its success is: will developers adopt open source (and specifically Meta) to build new products? Mark Zuckerberg shares his vision for the future in "Open Source AI Is the Path Forward" https://lnkd.in/eqbmx7_x?

27

1 Comment

AI startups are thriving 🤯 AI companies capture 28% of global funding in Q2’24. This is the highest quarterly share on record ❗ source: https://lnkd.in/geZD8xbf #venturecapital #vc #technology #tech #deeptech #startups #invest #artificialintelligence #ai #business #funding #cbinsights

10

We’re getting to that point in a technology’s journey where we open or keep closed source. Meta just released Llama 3.1, the largest-ever open-source AI model, which apparently outperforms GPT-4o and Claude 3.5 Sonnet on several benchmarks. MZ says that Meta AI will be the most widely used assistant by the end of this year! Interesting times ahead for what is set to be the biggest tech advance since the internet. Where do you sit in terms of open and closed source and does the same rule apply for AI? #ai #genai #chatgpt #metaAI

11

1 Comment

It's time for SuperAI in Singapore so there's no better time for me to talk about this. Reid Hoffman recently put out a couple of videos of himself talking to an AI version of himself. It was an amazing technical achievement. Not only was the AI video indistinguishable from real footage, AI Reid was indistinguishable from real Reid! All of this cued breathless extrapolations from commentators: Feed in Isaac Newton's entire corpus and you can create an AI Newton whom you can ask scientific questions. Send an AI version of yourself to attend meetings on your behalf and give you a summary of the meeting when it ends. Or perhaps the most unusual suggestion I came across was to talk to one's own AI avatar as a form of self-therapy! All of these ideas seem to miss the basic essence of how these avatars are created: from codifiable, historical data. This means that these avatars can't incorporate new information. AI Newton has no idea about the theory of relativity, no matter how smart Isaac1.0 himself was. 🤷‍♂️ They also make the same mistakes that the original made: ask AI Newton whether lead can be turned into gold and it (he?) would say Yes. 🤔 Creativity is a mixed bag. Yes, they can create a painting of a cat 🐱 in the style of Monet, but is that really creativity? Is it more akin to applying a sepia filter on a photo to make it look old-timey? More subtly, we as humans are continuously shaped by random events and luck. Generative AIs don't have access to random influences from the future – their training data stops when it stops and the avatars they create become outdated models with every passing day. Even if you only care about the past, your model is only as good as its inputs, and your input data is far from complete. Who we are as humans isn't defined only by our social media posts and other codifiable information – it also depends hugely on other facts we know but never write down, our moods and even our physical bodies. Eleven-year old Murli was definitely Murli. But he didn't write down everything he thought, he didn't know the things today's Murli does, and he was a little boy who didn't have middle-aged aches and pains 😆 And sixty-year old Murli will think differently again from today's Murli. What do you – not your AI avatar – think?

34

6 Comments

Ever expanding context window (Google Gemini has a huge lead here) is making prompt engineering one of the most effective and cheapest way to significantly improve LLM’s output. #GenAI #PromptEngineering #ContextWindow #Prompting

7

1 Comment

Is generative AI the next big thing or just another tech bubble? The hype is real, but so are the concerns. Academics warn of inflated promises and overvalued AI stocks, drawing parallels to the dot-com boom. Yet, some analysts remain optimistic about AI's long-term potential. As we navigate this AI-driven landscape, the key lies in distinguishing between genuine innovation and speculative fervor. #AIBubble #TechInvestment #FutureTech #AIEthics #InnovationHype What's your take on the current AI hype? Share your thoughts below.

5

Introducing Captable Simulator - a simple tool for founders to simulate their captable over subsequent rounds of funding. Try it here: https://lnkd.in/dnqxegFp This tool was built using Anthropic's Claude Sonnet 3.5 in a few days. Tips for coding with Claude: 1. Use it to ideate before code gen. It comes up with a better process, better feature reccos & even generates SVG mockups 2. be as descriptive as you can in your prompts. give it a lot of context 3. ask it to be brief & helpful in its responses 4. think of all features before prompting Claude 5. In the long chat, code gen would get RL'd at around 350-400 lines of code for me. So generate only updated code blocks 6. if you get RL'd, new chat windows don't have memory persistence. You'll get entirely different code. The Big Question I wanted answered with this experiment - will GenAI replace CTO's? The short answer: Not today for sure. In the future, maybe. Learnings:  - Claude like tools are excellent at 10x'ing or even 100x'ing a coder's capability. It's definitely not a replacement. - A paid Claude account feels like a superpower. Anything I could think of, Claude would conjure.   - You still need someone who understands code to get into the code. It's very easy to mess up even generated code. - Gen code is not perfect. I'd often run into redundant code - For an early stage startup, it should be imperative to use Claude for rapid prototyping and user testing. For larger enterprises, it's difficult to build entire products with GenAI (IMO) because of the very real rate limits. - for another app, I tried giving Claude a screenshot of a paid app I use and told it to generate the code to replicate its features. Claude fumbled and didn't give me what I wanted. So, no, apps will not be easily replicated with such tools, yet. - However, none of this is to say it's not impossible to think that, directionally, Claude could potentially replace entire coding teams, it could code entire products from scratch, it could replicate competitor products or features with just a screenshot. Just not today.

24

2 Comments

🅷🅾🆆 🆆🅸🅻🅻 🆆🅴 🅴🆅🅴🆁 🅺🅽🅾🆆 🆆🅷🅰🆃’🆂 🆁🅴🅰🅻? Saw a great post by Stephen Tan on the necessity of AI needing guardrails... whilst I don't disagree, the issue is that the authorities have shown for decades that they lack the expertise to navigate tech. Who is capable to monitor and police the advances that are happening with the exponential growth in the ability of technology to recreate the world. This is truly fascinating territory as societies head into this maelstrom of potential chaos. This video is fun and wild and a lot more tasteful than seeing Trump kiss Kamala... none of us need to see those images. So we are in the 2nd half of 2024 and it is racing away from us. The news cycles are chaotic and that notion of simulation theory just seems a little more plausible by the day. 😃 🤣 😯 The guardrails that defined 'normal' seem to be out of action. Whatever is coming, is coming so it's really a matter of how to adjust and prepare for an entirely new paradigm. Here's a fun acronym to help you prepare... #FARQVARD - sure it's not really the best acronym but depending on how you pronounce it, it sounds amusing. F🅛🅔🅧🅘🅑🅛🅔 A🅖🅘🅛🅔 R🅔🅢🅘🅛🅘🅔🅝🅣 Q🅤🅘🅒🅚 V🅔🅡🅢🅐🅣🅘🅛🅔 A🅓🅐🅟🅣🅘🅥🅔 R🅔🅢🅟🅞🅝🅢🅘🅥🅔 D🅨🅝🅐🅜🅘🅒 However you want to describe it, these attributes are going to be essential in the world we are creating. In the midst of all this digital chaos, it's easy to forget that some of the best answers to navigating our rapidly changing world lie in the simplest of things—like getting back to nature. My walks, both alone, with friends and with clients in MacRitchie reminds me of the balance we all need. The calm of the forest and the rhythm of the trails offer a clarity that no technology can replicate. As we all navigate this new paradigm, let’s not forget to reconnect with nature—and with each other. Walking and talking, we realize that we’re all on this journey together. Through community, conversation, empathy, and understanding, we find resilience and clarity. If you're seeking purpose or direction, reach out—sometimes the answers are closer than we think, just waiting for us to pause, listen, and connect.

55

17 Comments

🚀 28 AI startups raised $100M+ in 2024. What's their secret sauce? Key lessons for founders: 1️⃣ Solve big problems Xaira Therapeutics raised $1B for AI drug discovery, while FundGuard secured $100M for AI-powered investment accounting. The message? Focus on industries with complex, high-value problems where AI can make a significant impact. 2️⃣ Build on existing tech CoreWeave raised $1.1B for specialized GPU infrastructure, and Bright Machines secured $106M for AI-driven manufacturing robots. Look for bottlenecks in current AI implementations and innovate there. 3️⃣ Pioneer new frontiers Suno raised $125M for AI music creation, while EvolutionaryScale got $142M for biological AI models. Don't be afraid to explore uncharted territories where AI can create entirely new possibilities. 4️⃣ Focus on enterprise solutions AlphaSense raised $650M for AI-powered market intelligence, and Glean secured $203M for AI-driven enterprise search. Remember, enterprises are willing to pay for AI that delivers clear ROI. 5️⃣ Develop AI infrastructure Etched AI raised $120M for AI-specific chips, while Scale AI got $1B for data labeling services. The AI boom needs better tools and infrastructure. Can you provide them? 💼 For VCs: Look for startups tackling real-world issues or creating new markets. Evaluate the potential market size and the startup's ability to capture it. Consider the team's expertise in both AI and the target industry. Assess the scalability and defensibility of the AI solution. And don't ignore "unsexy" sectors that could be transformed by AI. The AI gold rush is on, but the biggest wins are going to those who can demonstrate real-world impact, scalable solutions, and visionary thinking. Whether you're a founder or a VC, the key is to look beyond the hype and focus on tangible value creation. What's your take on the AI startup landscape? Are you seeing similar trends in your area of focus? Let's discuss in the comments! 👇 #SkimAI #EnterpriseAI #Founders #CEOs #Startups #AIFunding #AIandYOU

8

1 Comment

🚀 Why Prompt Engineering is a Great Way to Test Drive GenAI 🚀 Are you considering diving into the world of Generative AI (GenAI)? Start with Prompt Engineering! Here's why: Low Barrier to Entry: Crafting prompts is straightforward, allowing you to quickly explore GenAI capabilities without deep technical expertise. Immediate Feedback: Instantly see how GenAI responds to different prompts, helping you understand its potential and limitations. Cost-Effective: Experimenting with prompts requires minimal resources, making it a budget-friendly way to pilot GenAI. Customizable Outputs: Tailor prompts to your specific business needs, enabling you to fine-tune the AI's responses for maximum relevance. Scalability: Start small with prompt engineering and scale up as you gain confidence and insights. Kickstart your GenAI journey today with prompt engineering and unlock innovative possibilities for your enterprise! 🌟 #GenAI #PromptEngineering #AI #TechInnovation #EnterpriseAI #AIExperimentation #FutureOfWork #NIMBL #EKAI

4

How did OpenAI go from a $1 billion bet to a $90 billion AI leader? OpenAI's story, with 8 rounds totaling $11.3 billion, is a prime example. Let's unpack its funding strategy timeline: 𝟮𝟬𝟭𝟱: 𝗔𝗻𝗴𝗲𝗹 𝗜𝗻𝘃𝗲𝘀𝘁𝗼𝗿𝘀 𝗟𝗶𝗴𝗵𝘁 𝘁𝗵𝗲 𝗦𝗽𝗮𝗿𝗸 > OpenAI won a $1 billion grant from tech titans Elon Musk and Peter Thiel, creating a robust runway. 𝟮𝟬𝟭𝟲-𝟮𝟬𝟭𝟴: 𝗡𝘂𝗿𝘁𝘂𝗿𝗶𝗻𝗴 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝘄𝗶𝘁𝗵 𝗣𝗿𝗲-𝗦𝗲𝗲𝗱 𝗦𝘂𝗽𝗽𝗼𝗿𝘁 > Nonprofits like the Ethereum Foundation and Y Combinator provided early funding ($25 million + $120,000). 𝟮𝟬𝟭𝟵: 𝗔𝗱𝗮𝗽𝘁𝗶𝗻𝗴 𝗳𝗼𝗿 𝗚𝗿𝗼𝘄𝘁𝗵: 𝗧𝗵𝗲 𝗙𝗼𝗿-𝗣𝗿𝗼𝗳𝗶𝘁 𝗣𝗶𝘃𝗼𝘁 > Desperately needing 𝘮𝘰𝘳𝘦 funding but lacking profitability, OpenAI created a for-profit division to attract investors. > It capped potential profit earnings, and CEO Sam Altman had no equity stake. > This attracted a $1 billion Series D investment from Microsoft. 𝟮𝟬𝟮𝟬: 𝗘𝗮𝗿𝗹𝘆 𝗠𝗼𝗻𝗲𝘁𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗣𝗮𝘃𝗲𝘀 𝘁𝗵𝗲 𝗥𝗼𝗮𝗱 > OpenAI sold ChatGPT access to generate revenue and partnered with businesses like Reddit and Salesforce. > While not yet profitable, it established a revenue stream ($30 million). 𝟮𝟬𝟮𝟭: 𝗘𝘅𝗽𝗮𝗻𝘀𝗶𝗼𝗻 & 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝗶𝗰 𝗜𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁𝘀 > OpenAI broadened its reach with ChatGPT-4. > It created a venture fund for startups using its technology. > A $100 million donation from Vitalik Buterin further fueled growth, reaching $36 million in ARR. 𝟮𝟬𝟮𝟮: 𝗜𝗻𝗻𝗼𝘃𝗮𝘁𝗶𝗼𝗻 𝗔𝗺𝗶𝗱𝘀𝘁 𝗜𝗻𝘃𝗲𝘀𝘁𝗺𝗲𝗻𝘁 > OpenAI continued pushing boundaries by launching ChatGPT-5 and OpenAI Codex. > Although its ARR increased to $200 million, it still made a $540 million loss. 𝟮𝟬𝟮𝟯: 𝗧𝗵𝗲 𝗣𝗿𝗼𝗳𝗶𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗧𝗶𝗽𝗽𝗶𝗻𝗴 𝗣𝗼𝗶𝗻𝘁 (𝗖𝗿𝗶𝘁𝗶𝗰𝗮𝗹 𝗖𝘂𝘀𝘁𝗼𝗺𝗲𝗿 𝗠𝗮𝘀𝘀) > OpenAI achieved a breakthrough after 7 years, exceeding $2 billion ARR. > This 900% ARR growth made it a self-sustaining AI leader. > Resulting in a $10.305 billion Series E investment, pushing its valuation to $90 billion. Key Takeaways: 📈 𝗜𝗻𝗰𝗿𝗲𝗺𝗲𝗻𝘁𝗮𝗹 𝗚𝗿𝗼𝘄𝘁𝗵 𝗕𝗲𝘁𝘄𝗲𝗲𝗻 𝗙𝘂𝗻𝗱𝗶𝗻𝗴 𝗥𝗼𝘂𝗻𝗱𝘀 > Each round helped OpenAI scale and achieve specific milestones. 📈 𝗗𝗶𝘃𝗲𝗿𝘀𝗶𝗳𝗶𝗲𝗱 𝗙𝘂𝗻𝗱𝗶𝗻𝗴 𝗦𝘁𝗿𝗮𝘁𝗲𝗴𝘆 > It leveraged venture capital, angel investors, donations, grants, and strategic partnerships. 📈 𝗣𝗿𝗼𝗳𝗶𝘁𝗮𝗯𝗶𝗹𝗶𝘁𝘆 𝗥𝗲𝗾𝘂𝗶𝗿𝗲𝘀 𝗖𝗮𝗹𝗰𝘂𝗹𝗮𝘁𝗲𝗱 𝗥𝗶𝘀𝗸 > OpenAI monetized between funding rounds. Code, culture, or cash? What is the primary driver of OpenAI's financial success? Credit: CB Insights, Tracxn, Sacra, & Analytics Insight. PS 🔔 Follow me for strategies and resources for startups and VCs! Looking to raise capital? 💸 https://foundersuite.com/ for startups 💸 https://fundingstack.com/ for VCs and investors At only $250/month: Fundingstack gets you access to 267K+ global investors. Or you can send me a direct message with "FREETRIAL" and I'll give you a secret access link.

50

1 Comment

🍓Exciting Times in AI with the New OpenAI Chat-GPT Model (Codename: “Strawberry”)!🍓 The world of AI just got a major upgrade with the release of the latest Chat-GPT model—codename “Strawberry” (o1). 🍓This model is pushing boundaries, incorporating data pruning and process supervision techniques to accelerate learning and improve model efficiency like never before. 🔑 Here’s why this is a game-changer: - Data Pruning: By selectively focusing on the most valuable data, the model achieves better performance with less computational power. This not only reduces resource consumption but also unlocks better-than-power-law scaling. - Process Supervision: Instead of only checking final answers, the model gets feedback at every step, making it more reliable in multi-step reasoning tasks—like complex problem-solving or decision-making. This is especially important for large-scale models, enabling more robust outputs. 💡 The combination of these techniques means more efficient AI that’s smarter, faster, and less resource-intensive. This is the future of AI models—efficient, powerful, and ready to scale across industries! Curious about how AI is evolving? Let’s dive deeper into this revolution! 🔍 #AI #MachineLearning #ChatGPT #Strawberry #DataPruning #ProcessSupervision #LLM #NeuralNetworks #Innovation Papers: 📝 “Beyond Natural Scaling Laws” https://lnkd.in/eZge7PfM “Let’s verify step by step”https://lnkd.in/eF3MhuYF

1

OpenAI today announced their new GPT-4o model. Here are the key points you should know: Omni: The "o" in GPT-4o stands for Omni. OpenAI is now focusing on a cross-functional user experience, integrating text, audio, and images into the same AI pipeline. Unlike previous versions of ChatGPT, where you had to convert images into prompts or text into voice, the Omni model offers a seamless experience. This results in faster and more intuitive interactions. Improved Performance: GPT-4o is faster, a LOT FASTER, and more capable across many tasks. It outperforms previous versions of GPT-4, as well as popular models like Gemini and Claude, in numerous publicly listed tasks and test comparisons. While further research will validate these claims, the progress across text, audio, and other AI applications is impressive. When you don't have to wait, you can iterate on prompts and answers much more quickly. Desktop Version: OpenAI is releasing a desktop version of GPT-4o that can access the context of your desktop and screen. This feature enables various collaborative scenarios where ChatGPT understands the application in use. This is perhaps what Clippy was always trying to be, only this AI is exceptionally more helpful. Another significant development is that AI computation is now happening both on-device and off-device. This is expected to be further enhanced by upcoming technology from Apple, likely through next-generation assistants. Additionally, streaming camera and audio input allows for fluid conversations, interruptions, and continuations without losing context. Despite the technology being only a day old, my initial experience over the past few hours indicates that while the AI revolution is well underway, the UX revolution driven by these technologies is just beginning.

31

2 Comments

Many AI startups seem to be building their products on a hope and a prayer! They have surrendered control of core workflows in their businesses to 3rd-party LLM APIs. These startups communicate with these LLMs via special incantations that are expressed in arcane structures (called prompts but more like pleas), cajoling, negotiating, hoping that the LLMs would grant their wishes. Prompt Pundits on social media and blogs argue about the best way to encode the needs of the business into these entreaties, just so that the LLMs would bless them with consistent results. They offer their services to these startups as prompt engineers/consultants, but they aren't able to control the whims of LLMs either (let alone updates, retraining schedules of LLM providers). LLMs operate in mysterious ways, indeed! I'm exaggerating a bit here, but not by much. Imagine running a tech startup on a cloud provider that executes 75% of the code, and a different 75% each time. No Eng Leader would ever allow this; but many seem to be ok with this type of execution substrate for LLM-powered workflows. Don't get me wrong. I'm a big fan of LLMs; they have enabled me to build AI features at Bluebirds with an accuracy and iteration velocity that I could only dream of during my last startup. So, if you want to have more control over your LLM-powered workflows, here are some recommendations. 1. LLMs are best at extraction from unstructured text and generation of natural language. As much as possible restrict calls to LLMs to these steps. 2. As a Product Leader you know the job-to-be-done for your users best; if you don't, this is the what you should spend time on. You cannot rely on LLMs to fill this gap! 3. Break down the workflows into as granular steps as possible, and enlist LLMs for the parts they are best at. (see pt 1) 4. If you must use LLMs for reasoning across multiple steps, then try providing the LLM with the explicit flows/logic. For some domains (like Sales) there is a lot of content online that LLMs have been exposed to. This helps them create generic plans ("chain of thought!"), which is sufficient if you wanted to offer your users as much insight as a click-baity sales blog post. But if you have unique insights into the workflow (see pt 2) then break it down into smaller steps and use LLMs strategically for them. This is your moat, after all! This is the path we have taken at Bluebirds when designing our Agentic systems. Curious to hear if others have more or different recommendations.

58

11 Comments

OpenAI launching their speech-to-speech real-time voice API is a big deal. It's now possible to build experiences like Advanced Voice Mode. Go build with OpenAI and LiveKit! Here's why the speech-to-speech API matters so much. Historically models could only accept a text input, and only give a text output. In order to build a voice experience you needed to string together multiple tools. You'd use something like LiveKit to handle all of the network infrastructure and optimization. Then you'd use a speech to text provider to convert the audio file to text to feed into the model. The model runs inference and then spits out text, that is feed into a text-to-speech provider. That audio file is then sent over the network (LiveKit) and played out the speaker in the application. This creates two "issues" 1) All of that conversion takes time. Stringing together multiple processes just increases the lag / wait time from one speaker to the other 2) Converting an audio file to text is a form of compression. And when you compress a file you loose data. There's important context in an audio file that goes away in text. The tone of voice, which words are emphasized, etc These two "issues" make it harder to create human like experiences. Either they aren't real time (because so much extra work has to happen in the background), or they lack a true emotional connection when it's all text based OpenAI launching the speech-to-speech API is a huge first step in creating truly human-like experiences with AI And when it comes to the network infrastructure - many builders will first try web sockets or vanilla WebRTC. It's a good way to start, but as you scale, performance issues will be felt. Instead, start from a solid foundation with LiveKit

67

1 Comment

Founders - Here are the 10 most active VC investors in generative AI. 📈 Sequoia Capital: One of the world’s largest VC firms, has invested in some of the most prominent AI startups like ChatGPT creator OpenAI and legal AI assistant Harvey. Andreessen Horowitz: Andreessen Horowitz has gone all-in on AI, pledging to back as many startups in the space as possible. Its portfolio includes Pinecone, an AI database and application development startup, and large language model creator Mistral, the French rival to OpenAI. Pioneer Fund: Investing exclusively in Y Combinator alumni, Pioneer Fund’s portfolio includes BoothAI, an automation startup for e-commerce, and Decohere, an image and video generation company. The firm has a specific AI and ML fund that is run by technical AI founders. Soma Capital: Soma Capital invests in startups across all verticals and stages with offices around the world. Its AI portfolio includes Cognition Labs, an AI agent developer, and Imbue, an AI systems creator that reached unicorn status in September. Khosla Ventures: Khosla Ventures founder Vinod Khosla has been a huge champion of AI and was the first VC backer of OpenAI. The firm has backed AI startups across a wide spectrum of areas like Replika, an AI chatbot companion startup, and SketchPro, a design assistant geared toward architects and designers. Alumni Ventures: Operating university-specific funds across the US, Alumni Ventures has an AI and machine learning fund and has backed some of the biggest startups in the space. Its portfolio includes the large language model developer Cohere as well as Lamba, a GPU cloud computing startup, both of which have achieved unicorn status. Lightspeed: Lightspeed has invested in some of the most prominent AI startups on the scene. Its portfolio includes Stability AI, the creators the video generation model stable diffusion as well as Snorkel AI, an LLM training platform for the enterprise. Gaingels: With a mission to promote social change and inclusivity through investing, Gaingels has backed startups like People.AI, an automation platform for e-commerce providers, and Shield AI, which is developing AI pilots. Tiger Global Management: Tiger Global had made some big bets in generative AI, backing two of the leading large language model creators, OpenAI and Cohere. The firm has also invested in OctoAI, an orchestration startup for tuning and tweaking models. Greylock Partners: Greylock was founded in 1965 and has backed generative AI startups including Tome, an AI storytelling platform, and Inflection, an AI chatbot creator that was absorbed by Microsoft in an unusual $650 million deal in March. (𝘍𝘳𝘰𝘮 𝘰𝘶𝘳 𝘧𝘳𝘪𝘦𝘯𝘥𝘴 𝘢𝘵 𝘗𝘪𝘵𝘤𝘩𝘉𝘰𝘰𝘬.) There you go! Need help, reach out. ~~~    Like this? Follow Gerald Duran for daily Startup and VC advice, and sometimes — a kick in the nuts. #startups #founders #venturecapital #entrepreneurship 

74

47 Comments

A venture studio 110X return of $110M without raising outside capital (...steal this model): Most new venture studios immediately try to raise outside funding to get started. But there's a smarter way to bootstrap your studio to profitability: Build your core operating companies FIRST. Here's how it works: 1️⃣ Identify the key services every company you build will need (dev, design, marketing, HR, etc) 2️⃣ Launch these services as standalone, profitable agencies 3️⃣ Use the profits to fund your studio operations 4️⃣ Offer these services to your portfolio companies at cost or free, subsidized by profits from the agencies 5️⃣ Raise outside funding in to the portcos AFTER you have traction and proof points Why does this work so well? ✅ You reach default alive with infinite runway first ✅ You maintain complete control of your studio ✅ You build the assets your portfolio companies need most ✅ You generate revenue first ✅ You can raise capital on your timeline and terms for your studio and your portcos I recently studied a studio in Dubai that used this exact model. They started with a dev agency, then expanded their portfolio. 11 years later? Their initial $1M is now worth $110M — a 110X return! That puts them in the top 1% of venture funds. No outside funding needed. The studio model is evolving. Creative funding is the future. Build your core assets first. Control your destiny. There are more ways than raising to start a studio. P.S. I am opening 3 more client slots for the remainder of the year. If you are interested in coaching, advice, and hands on help building your studio, DM me ‘VSAnew’ and we can explore working together.

105

49 Comments

Meta has positioned itself as the open-source hero in AI, with Mark Zuckerberg penning blog posts extolling open source as the way forward. But many in the industry have been skeptical, questioning just how “open” Meta’s Llama models really are—and as of yesterday, they have a new rallying point. The Open Source Initiative (OSI), a long-time steward of open source principles, jumped into the debate by releasing version 1.0 of its Open Source AI Definition (OSAID). Here are the 4 critical pillars for a model to be truly “open source” in AI: ♻ Reproducibility - Sufficient detail must be provided for others to meaningfully recreate the model. 🔍 Transparency - Clear disclosure of training data sources, processing methods, and licensing is required. 🔢 Full Component Access - Access to all components—training code, data filtering—is a must to grasp how it was built. ⚙ Usage Rights - Freedom to use, modify, and build upon the model without restrictions. This puts Meta’s “open” claims in a gray zone: the training data is hidden, the compute required for deployment is out of reach for most, and usage is restricted for platforms with over 700 million monthly active users.  Meta, of course, argues that “older definitions don’t account for today’s advanced AI complexities.” They make their models free and openly available, putting some restrictions in place to keep people safe - and will continue to do so regardless of ’technical definitions’. Realistically, I think there's a balancing act between the idealism of fully transparent, reproducible models and the realism of massive resources needed to develop them. Sometimes, a “not-quite-open” solution still delivers substantial value. Expect more debate over what qualifies as “open” in the coming months. Until then, the verdict is clear: a rose by any other name may smell just as sweet but Llama by any other name might not graze as freely. 🌹 🐐

99

13 Comments