Andrés Soler

𝗬𝗼𝘂𝗿 𝗘𝗹𝗲𝘃𝗮𝘁𝗼𝗿 𝗣𝗶𝘁𝗰𝗵 & 𝗣𝗶𝘁𝗰𝗵 𝗗𝗲𝗰𝗸: 𝗞𝗲𝘆𝘀 𝘁𝗼 𝗜𝗻𝘃𝗲𝘀𝘁𝗼𝗿 𝗦𝘂𝗰𝗰𝗲𝘀𝘀 Crafting a compelling pitch is essential for startups seeking funding. Here’s how to structure your approach for maximum impact: 1️⃣ Introduction: Begin with who you are and what your company does. Example: "I’m Jane, CEO of Wanoxy, a platform that integrates Google Docs across all devices." 2️⃣ Current Status: Highlight where you are now. Example: "We’ve launched iOS v2.0 and gained 2M users in 6 months." 3️⃣ Next Steps: Share your immediate plans. Example: "We’re launching our Android version next quarter." 4️⃣ Market Opportunity: Illustrate your market size and potential. Use data to show growth opportunities. 5️⃣ Milestones: Explain how you’ll use the funding to achieve specific goals over the next three quarters, such as product releases, team growth, and market penetration. 6️⃣ Team Strength: Highlight why your team is uniquely suited to execute this vision. 7️⃣ Financials: Present realistic, scalable projections that support your valuation. 8️⃣ The Ask: Clearly state how much you’re raising and how the funds will be allocated. Example: "We’re seeking $1M to scale operations and expand our platform." PS. check out 🔔 for a winning pitch deck the template created by Silicon Valley legend, Peter Thiel https://lnkd.in/ejp-Bhnu

37

4 comentarios

These tech companies are making so much money it's just mind-blowing 😳 Tech revenue per employee of a few hidden gems that are not on this list: - Tether.io $62 million - Valve $37.1 million - Telegram Messenger $5.2 million As Apple CFO Luca Maestri said, "If you have the right people you don’t need a lot of them”. Money printers. h/t Trung Phan P.S. for more interesting stuff, check out 🔔linas.substack.com🔔, it's the only newsletter you need for all things when Finance meets Technology. For founders, builders, and leaders.

11

1 comentario

Why does UI still matter in an AI-first world? 🚀 AI-first, agent-driven workflows sound great—let AI handle everything behind the scenes. But when something goes wrong, you need more than magic; you need transparency and control. 👀 In RevOps, imagine managing a deal desk workflow. An AI can automate deal routing and approvals—but when a big deal gets stuck, you're left wondering why. A clear UI changes that. It lets you see what’s happening, adjust rules, tweak paths, and own your processes. 💡 Simplicity is nice, but true impact comes when you give people the tools to adapt and make AI work for them. Visual development isn't just about building—it's about refining, experimenting, and really owning workflows. That’s why UI still matters. It’s what turns AI from a black box into a real partner. PS: This post was inspired by an awesome conversation I had with Sam Davyson from Stacker the other day.

9

4 comentarios

🔍 The Future of Solvers in Intent-Based Systems: Challenges and Solutions In intent-based systems, users define what they want to achieve (like buying or selling an asset), and solvers find the optimal way to execute these intents. However, this model presents significant challenges and conflicts of interest. The Core Problem: Conflicts of Interest Solvers, with access to advanced tools and data, may leverage user-defined intents for their personal gain instead of executing them as expected. With little oversight, solvers can spot lucrative opportunities (like arbitrage) and use private tools and wallets to act on them, bypassing the user’s original intent. Why the Current Model is Unsustainable Misaligned Incentives: Solvers prioritize their profits over the user's intentions, especially when acting in their interest yields lower rewards. Short-Lived Opportunities: In competitive markets, profits from strategies like arbitrage are fleeting. This encourages solvers to act quickly and independently. Competitive Tools and Secrets: Solvers, driven by the need to maintain an edge, rarely share their tools or insights with other participants, further complicating trust and collaboration. The Solution: Separate Solvers from Activators We need to rethink the system structure to align incentives and ensure fair execution: Transform Solvers into Activators: Activators are rewarded for executing intents, not based on the value of the transaction, similar to how miners are compensated for mining blocks. Users as Their Own Solvers: Users can set up personal bots (e.g., via Telegram) to monitor market conditions. When conditions align, the bot signs the transaction, and the activator executes it on-chain. Transparency and Control: Users retain full control over the process, ensuring no conflict of interest. The system becomes more transparent, fair, and efficient. Example Use Case A user sets an intent: Buy on Binance and sell on Uniswap if there’s a $5,000 arbitrage opportunity on $100,000 volume. The user’s bot monitors the market. When the conditions are met, the bot signs the transaction, and the activator executes it. Regulatory Challenges – An Unexplored Topic Does a solver performing currency exchange need a license? Is KYC (Know Your Customer) mandatory for solvers to comply with regulations? Additionally, development challenges remain—how can solvers efficiently convert intents into operations that succeed under real-world conditions? If there's interest, I'll explore these challenges in future posts. Let’s design systems that promote trust, fairness, and decentralization. Intent-based systems hold tremendous potential—but only with the right incentive structures and operational models. #DeFi #Blockchain #IntentBasedSystems #Arbitrage #Decentralization #Web3

2

Google's recent earnings call confirmed AI is not just hype — it's driving real results. In particular, Chief Business Officer Philipp Schindler stated “AI really supercharges search. Our new AI-powered features make searches more helpful, and we continue to see great feedback.” As Google reaps the rewards of AI in #search, retailers and brands can too! How to win in this AI-driven landscape: * Optimize Product Details: Ensure your products are discoverable with rich, relevant data expressed using the natural language of your customers. * Leverage AI Too!: Utilize vertical-specific AI-powered platforms to enhance your product listings and ad strategies. * Stay Agile: As always, continuously refine your approach based on AI-driven insights. Remember, Google's search advancements mean improved opportunities for you to connect with consumers this #Holiday season. By optimizing your products and leveraging AI to do so, you can improve your Google ad spend effectiveness and claim your share of the AI-driven revenue boom. The best part? These same AI strategies benefit product search and discovery beyond Google as well. Don't watch from the sidelines—start integrating AI into your retail strategy today! #RetailAI #Ecommerce #GoogleAds #RetailTech

86

3 comentarios

The Top 12 Small Language Model Startups | Many businesses don't need huge proprietary Large Language Models (LLMs) for the vast majority of their use cases. Small Language Models (SLMs) are less resource-intensive than LLMs (like GPT-4) and offer several advantages: 💸 Cost-effectiveness ⏱ Faster deployment times, and 💻 Reduced computational requirements. S/O to Adya Pandey, Intelligence Analyst at CB Insights for her recent 2Q24 State of AI presentation | | #genai #ai #generativeai #artificialintelligence #business #ceo #technology | Hugging Face Sakana AI Sarvam AI Predibase Arcee.ai Nomic AI PremAI Malted AI Argilla Nous Research Claro AI Glaive |

55

7 comentarios

From Mixtral to Grok, MoE is scaled well to serve tasks of small to mid-model size. If you are planning to build MoE's, it'd be great to join the paper presentation by Niklas. https://lu.ma/yzxim9wd

7

GPT-4o Mini was released this week - should your firm get access? Let’s dig into it. GPT-4o Mini is a compact version of GPT-4o, offering performance nearly on par with GPT-4 Turbo, the premium model of the past year. OpenAI’s new model is outperforming every other model on the market (it’s number 2 on the LMSYS leaderboard, second to only GPT-4o! 🤯) This could be why Meta also announced the release of Llama 3.1😉 So what does this mean for the future of legal tech? Expect tools and applications powered by LLMs to continue generating high-quality output with a significant cost reduction. This could be a game changer for your firm’s GenAI strategy. There’s a LOT of information out there. So, to avoid decision paralysis whenever a new LLM is announced, here’s a five step framework to assess these models as they land. STEP 1: Review Benchmarking of Smaller Models GPT-4o Mini stands out among smaller LLMs like Google’s Gemini Flash. It offers comparable benchmarking to its larger sibling, GPT-4.o, at a fraction of the cost. It scored 82% on MMLU and outperforms GPT-4 in chat preferences on the LMSYS leaderboard.✅ STEP 2: Assess API Availability The GPT-4o Mini API is readily available via your existing Azure subscription, eliminating the need for approval rounds or new applications. 💪 STEP 3: Test the Output Quality Yourself After evaluating the accuracy, precision, and sensitivity through public benchmark reports, test the general model yourself. Assess how well the LLM performs on tasks you understand. You can do this through personal testing or by consulting your GenAI partner. STEP 4: Examine Context Window Size GPT-4o Mini has a context window of 128,000 tokens. This allows the model to hold long conversations and process large documents so it's ideal for legal work. 🤝 STEP 5: Assess Performance against Cost The new model is half the price GPT-3.5 Turbo. Consider this price reduction when planning your innovation strategy. Just as your firm invests in high-performing partners based on cost-benefit analysis, apply the same logic to LLMs. 💰 TLDR: GPT-4o Mini offers nearly on-par output to GPT-4o for a fraction of the cost AND is outperforming every other model on the market (see comments for a link to the LMSYS rankings). And it’s only going to get better from here. Have you had a chance to play with GPT-4o Mini or another recent model? LMK below!👇 _ PS: More on Llama’s 3.1 model coming soon - looking forward to seeing the benchmarks! 👀 PPS: The SpringLaw platform is LLM agnostic, making it easy to add GPT-4o Mini in to your firms LLM stack! If you want to create bespoke tools and client solutions, DM me today to get started. #MakeGenAIWorkForYou #GPT4oMini #GPT4 #GenAI

12

1 comentario

This graph from Google Trends compares Wooclap (in blue 🔵) to one of our competitors (in red 🔴). This competitor raised over $75 million. Their strategy mostly relied on aggressive and expensive commercial methods. We chose a completely different approach, focusing on the product and its virality: • Co-creation: developing every feature hand in hand with teachers, universities, and L&D managers to meet real-world needs. • User-friendly: making simplicity our core value. We believe an intuitive and easy-to-use solution is far more valuable than the buzz around "innovative" features that fail to deliver on their promises. • Business Model: offering a free version to amplify the product's virality, rather than relying on a closed system that requires talking to a sales representative for access. It wasn’t always easy, but it looks like our approach is paying off!

201

7 comentarios

Vendors and product owners: The FOMO is strong with generative AI features and the pressure to integrate an LLM (or something adjacent) into your product just to check the box has been intense for a while now. The result isn’t what you’re afraid of: there’s growing pushback from users whose workloads have *increased* with the influx of shoehorned AI features. More orgs are coming out to say they just aren’t seeing the benefits. It’s entirely reasonable to say, “hey, we’re doing some cool stuff with this!” but if you’re hunting for ways to turn GenAI into compelling features, you’re looking at it backwards. Look at your backlog and the requests your users already make: if there are easy (and cost effective) ways you can knock out those problems using stuff from the “AI Toolbox,” that’s great! But AI is not in and of itself a compelling product or even a compelling feature for most end users. In fact, there’s growing evidence that it’s perceived as a *negative* trust signal for your product. LLMs are a cool breakthrough in machine learning + natural language processing! But they cannot and will not ever do the hard strategic work for you.

142

22 comentarios

The growth chart for B2B SaaS from Paddle looks like dying aftershocks from a huge earthquake that was 2021-2022 for the industry. It's still quite hard to believe in a big bounce-back in 2024, and in this downturn, founders struggle to find new levers. However, the growth rate still affects acquisition valuations, and we'd like to share what exactly is being assessed on this front. Join us on May 2nd with Tim Heicks and Shiv Narayanan to talk about marketing due diligence in B2B SaaS acquisitions. Register with the link in the comments and prepare your questions 👇🏻

31

4 comentarios

Every early AI platform is going to run a pilot. Here are 10 questions you need to answer. So you got the pilot, now what? When you try out an AI SaaS product with customers, you need to figure out the big questions. Here are some of the most important ones: → What is the main problem your AI solution is addressing for the customer? You're solving a real customer problem. Make your product's benefits clear by knowing their struggles. → How easy is your product to use and integrate into existing workflows? Make your product easy to use. Test its interface to find and fix stuck spots. → Did the client help quantify how your AI helps them? Show how your product saves time, cuts costs, or improves key areas. → Are there any missing features or capabilities that customers need? Find what holds people back. Use this feedback to guide your next steps. → How does your solution compare to existing tools or processes? Compare your AI product to existing solutions. This includes manual methods and software. Find what sets it apart. → What level of customization or integration do we need? Check if customers need a custom fit. This impacts scaling. → How does your AI handle data privacy and security concerns? People have concerns about data usage, model training, and following rules. Being open helps win over customers. → What is the customer's willingness to pay for your solution? Find a pricing model that aligns with point #3. → How well does your AI perform in real-world scenarios? This is the scary one. Be open about things. → What level of ongoing support or training do we need? Do NOT flip on the switch and disappear. You need a training plan. This is how you'll learn a lot, make your product better, and figure out the best way to sell it. Did you cover all the bases - yes or no?

6

90% of all AEs use chatGPT wrong. Here’s how to truly unlock chatGPT for AEs with simple (voice) prompting: Imagine creating a Disco, Demo, and Closing checklist in under 3 minutes based on the best-in-class examples? So much GTM knowledge is shared daily by experts. But how to make it practical and consumable? I follow over 75+ amazing GTM subject matter experts to fuel my brain and Notion database with the latest strategic, tactical and operational advice on any Go-To-Market topic. The database currently consist of 4,300+ LinkedIn posts since July 2022 and 300+ GTM podcast shows of which these are the hottest 10 GTM topics, I have consumed most content from: ✔️ Stakeholder Management: 🤝🌐 ✔️ Internal Collaboration: 👥💼 ✔️ GTM Leadership: 🚀📊 ✔️ Sales Navigator: 🧭💹 ✔️ CFO & C-Suite: 💼🏦 ✔️ Multithreading: 🧵🔄 ✔️ Prospecting: 🔍💡 ✔️ Negotiations: 🤝⚖️ ✔️ Sales Demo: 🎥💼 ✔️ Champions: 🏆🌟 These are the top 10 Sales experts I have consumed most content from: 👉 Nate Nasralla - Selling with Buyers & Multithreading. 👉 Jason Bay - Turn Strangers into Paying Customers. 👉 Kevin "KD" Dorsey - Sales Leadership Coaching. 👉 👩🏻‍🏫 Krysten Conner - Sell More with Less Stress. 👉 Jen Allen-Knuth - Defeat Buyer's Status Quo. 👉 Kyle Asay - Sales Frameworks, AI & ChatGPT. 👉 Salman Mohiuddin - AE Sales Fundamentals. 👉 Chris Orlob - AE Sales Training & Coaching. 👉 Mike Gallardo - Pipeline & Closing Tactics. 👉 Brian LaManna - Prospecting & Closing. Simply ask any relevant GTM question to the database. Let me show you an example. These 3 prompts gave amazing results in minutes to create a checklist how to prep, run and follow up on your Discovery Call, Demo Call and Closing call. 𝗣𝗿𝗼𝗺𝗽𝘁 𝟭: "My team consists of Account Executives who are only focused on new logo acquisition. Their core job is handling inbound (direct booked in their agenda) and outbound requests (qualified by the SDR). So basically the mean themes of focus and execution are: ... I’m looking for best practices, tips, templates from the best content creators on ‘what good looks like’. Ideally i can turn this into a checklist per stage (see above) to upper the quality of execution by the team. " 𝗣𝗿𝗼𝗺𝗽𝘁 𝟮: "I want the tips outlined per each of the specific 12 stages outlined above. Provide per stage 3 best practices with reference to the content creator and the notion link." 𝗣𝗿𝗼𝗺𝗽𝘁 𝟯: "Extend each of the stages, each of the 12 stages, with an additional 4 examples. So, ideally, you create a 7-steps checklist per stage. So, in total, 7 steps for each of the stages, which is 7 multiplied by 12." The magic of chatGPT is in the specific prompts and how you follow up on the initial prompt. Want to see my 7 steps checklist for your next Discovery, Demo or Closing call? Comment '𝗚𝗧𝗠 𝗖𝗵𝗲𝗰𝗸𝗹𝗶𝘀𝘁' and I will share the link to the prompt AND the library.

429

757 comentarios

✨ The Google Mafia in AI: 14 Companies, $71.61B Valuation Fascinating discovery from our latest AI startup research: 14 of the top 50 AI companies are led by ex-Google employees, forming what we might call the "Google Mafia" in AI. The numbers are staggering: - $14.7B: Combined funding raised - $71.61B: Total company valuations - 28% of total funding among the top 50 AI companies [Full STUDY link in the comment - DM for details] Notable leaders and their ventures: ☑ Dario Amodei (Anthropic) - $40B valuation ☑Aravind Srinivas (Perplexity) - $8B valuation ☑ Arthur Mensch (Mistral AI) - $6.2B valuation ☑Aidan Gomez (Cohere) - $5.5B valuation What's remarkable isn't just the numbers, but the diversity of AI applications these leaders are pioneering: - Foundation Models (Anthropic, Mistral AI) - Search Innovation (Perplexity, Glean) - Enterprise AI (Cohere, Cresta) - Creative AI (Character AI, Pika) This pattern suggests Google has been an inadvertent incubator for the next generation of AI leaders. Their alumni are now shaping the future of AI across multiple domains. What are your thoughts on this trend? Is this the natural evolution of talent in tech, or is there something special about Google's AI DNA? ♻️ Let's celebrate these inspiring leaders. If you find this insightful, please share or repost to spread the findings. #AI #GoogleAlumni #Startups #VentureFunding #AIfunding

11

1 comentario

Deciphering the Challenges of Pricing LLM Models" 🧩💰🤔 ♻ Why is pricing an LLM model challenging? One of the most common LLM pricing models is "Token Based Pricing," where pricing depends on the number of tokens processed. Tokens represent words or subword units within the prompt sent to the LLM. However, companies are quickly realizing that this model has limitations, leaving significant revenue untapped. 📉 ♻ What are these challenges and limitations? ➊ Token Variability: Tokens can vary significantly in length and complexity, making pricing based solely on token count inaccurate. ➋ Context Matters: LLMs process tokens in context, meaning the surrounding tokens affect their meaning, requiring adjustments in pricing methodology. ➌ Subword Units: LLMs use subword units (e.g., Byte pair encoding or WordPiece), adding complexity to token counting and pricing. ➍ Ambiguity and Polysemy: Some tokens have multiple meanings, making it difficult to factor semantic richness into pricing. ➎ Language Considerations: Pricing based solely on token length becomes challenging when dealing with different languages. ➏ Customization and Fine-Tuning: The level of customisation and fine-tuning also impacts pricing, but quantifying these factors is currently difficult. As we enter the next phase of LLM adoption, with real use cases being developed, monetisation becomes a key focus. Similar to how "As-a-Service" models revolutionised cloud computing pricing, innovative pricing models are expected to emerge in the LLM domain in the coming months. 💡 #LLM #AI #PricingModels #GenAI

17

2 comentarios

💡Did you know that #AI is revolutionizing the way companies handle #CustomerService? 💬 From 24/7 support via chatbots 🤖 to hyper-personalized customer interactions, AI is driving a new era of customer satisfaction. In this AI Today episode, we explore real-world examples of how AI is improving customer service, boosting response times, and transforming the way businesses engage with their customers. 🎧 Listen here to learn more: https://lnkd.in/eMgYkADk #AIToday #podcast #customerservice #usecases

22