SEEKING FOUNDERS & TECH LEADERS
SEEKING FOUNDERS & TECH LEADERS VIEW MVL OPPORTUNITIES
VIEW MVL OPPORTUNITIES
Reflecting on 2023, a year of upheavals and advancements for tech startups, we witnessed a few significant shifts. Venture funding, which began to wane in 2022, continued its decline in 2023. Downturns are painful but also a natural part of economic and investment cycles that are out of our control.
Amidst the downturn, the rise of large language models (LLMs) like GPT-3, 3.5, and GPT-4 captivated us. Though gradual over a decade, their evolution still seemed sudden and remarkable. For those of us entrenched in machine learning and natural language processing (NLP), these models have seemingly solved many complex language understanding and translation challenges, an astonishing and groundbreaking development.
The accessibility of these new models and technologies is equally striking. Although a handful of large tech companies dominated the headlines, as usual, the open-source software (OSS) community has played a pivotal role, rapidly creating high-level toolkits that enable the development of innovative GenAI applications. This democratization of technology has been accelerated by the release of free, albeit less powerful, models by the OSS community, hinting at a future where the capabilities seen in the current generation of OpenAI models will become commoditized.
2023 was a whirlwind of exploration in LLM applications, sparking entrepreneurs and their ideas for new products and services. The ease of use of LLMs is notable; advanced degrees are unnecessary, as even basic Python skills suffice. This has leveled the playing field, allowing both large tech firms and small startups to innovate.
However, this democratization brings challenges to the startup ecosystem. The traditional tech startup playbook seems upended as LLMs alter the dynamics of competitive advantage and market entry strategies. This shift has prompted introspection within the startup community on adapting and reorganizing our strategy in this new landscape.
In the latter part of 2023, I had the opportunity to discuss these changes with prominent Seattle tech founders at the Madrona Labs Launchable event. Conversations with experienced entrepreneurs and investors Maria Karaivanova (WhyLabs), Diego Oppenheimer (Factory), and Jason Knight (OctoAI) revealed wisdom and insights into the evolving GenAI startup scene. Surprisingly, the core advice remained grounded in fundamental principles, though some aspects have gained new relevance post-2022.
In this post, I aim to distill our discussion on the essentials of the 2024 startup playbook, highlighting what has changed and what remains constant in this dynamic GenAI era.
A timeless principle in the seed stage of startup investment remains unchanged: investors are on the lookout for ventures representing massive opportunities led by teams with an “outsized” advantage to innovate and execute. An essential strategy for tackling these vast opportunities is to avoid trying to conquer everything at once, a mistake often called "boiling the ocean." Instead, success often lies in methodically addressing specific verticals one segment at a time. For instance, while your product and technology might have broad applications across diverse sectors like healthcare and financial services, selecting just one or even a part of one is crucial as your initial focus for go-to-market (GTM) strategy.
Launching a product or service is always a challenging endeavor. Gaining that initial traction and setting the wheels of progress in motion typically requires leveraging every available resource. Here, domain knowledge and experience become invaluable. It's not just about what you know but also who you know. Your professional network, an in-depth understanding of industry decision-making processes, and even the unspoken intricacies ("where the bodies are buried") can provide you with essential insights and leverage. This intelligence is particularly important in sectors with specialized data and information structures, like healthcare. Understanding these nuances can offer a significant competitive advantage, potentially saving substantial time and resources as you navigate the complexities of market entry and expansion.
No doubt, the emergence of LLMs has been a game-changer. I had plenty of "fall out of my chair" moments in 2023 experimenting with the technology. Their incredible capabilities are fascinating and will transform how we work and live.
However, the journey from a compelling technology to actual product market fit (PMF) involves multiple layers. While GenAI's capabilities are revolutionary, the real question is whether your product addresses an immediate need for your customer. Timeliness is critical, especially for startups with finite resources. It’s not just about where you envision the customer in the future; all of our panelists emphasized meeting them where they are now is crucial.
Equally important is the concept of "change readiness." Introducing a new product often necessitates behavioral changes, whether adapting to new data and capabilities or altering existing processes. Human nature and organizational culture can make this adaptation challenging. Therefore, your product design and go-to-market strategy need to address your customers' present state and anticipate their evolution over the next few years.
It's widely acknowledged that user-centric design is key, but faithfully implementing this in product development is more challenging than it seems. In a crowded and competitive field, usability, intuitiveness, and the delight your product brings to your users are more critical than ever. Achieving this requires ongoing refinement based on user feedback and genuine empathy for the user.
Navigating user experience development in GenAI-native applications presents unique challenges. We're all in new territory, and the transition between chat interfaces and traditional user experience (UX) design might see some oscillation before settling into some kind of hybrid model. I find it a bit ironic that while we've long imagined conversing with our technology in natural language (e.g., from sci-fi, “Computer: Warp speed engage!”), the reality is that many users are still adapting to chat interfaces. Many are unsure what to input and what to expect when the UX is a text field. Coupling this with the collective experience of a decade of woefully inadequate and frustrating chat technologies, this adaptation underscores the importance of meeting users where they are in their level of comfort with these new technologies and interfaces.
The journey with GenAI is a collective learning process. We need to unlearn less efficient habits and embrace more effective methods, a transition that will occur at varying paces across different user demographics and generations. Patience and understanding of these variations are crucial as we all adjust to the capabilities and potential of GenAI and the power our new products will bring our customers.
In my experience working with budding tech entrepreneurs, the topic of creating defensive moats is foundational. For the last 15 or more years, the most robust technology moats were built on proprietary, essential data, as machine learning ideas and technologies tend to commoditize rapidly.
Yet, with the advent of large language models (LLMs) that come pre-trained on massive data sets, the traditional data strategy and its relevance as a moat might be shifting.
Despite these changes, it's important to note that traditional machine learning is far from obsolete. Non-neural network models and the need for substantial labeled data remain crucial in many analytical applications. However, as we transition from proof of concept to deploying production-ready Gen AI applications, the emphasis might increasingly shift towards data quality rather than quantity. The era of “big data” has conditioned us to value quantity over quality data sets. But for the purposes of fine-tuning smaller domain-specific open source models (for both accuracy and costs), the future may favor smaller, high-quality data sets enriched with domain-specific insights.
The concept of a virtuous data cycle remains critical. Startups need to consider how their products, through customer interaction, generate valuable data that enhances their products, attracts more customers, and expands market reach. Building this cycle takes time, and startups are vulnerable in the early stages. However, a well-thought-out strategy around this cycle will be required from investors.
For as long as I can remember, the threat posed by tech giants has loomed large over entrepreneurs and investors. Questions like, “...so why wouldn’t Google, Amazon, Microsoft, Meta, or Apple just do this…?!” have always been common. But recently, this fear seems amplified, possibly due to the astonishing capabilities of Gen AI technology and the advantages big tech companies hold, such as access to cutting-edge tech and extensive user datasets.
However, an important perspective to consider is the agility of startups compared to the often slow and bureaucratic nature of large companies. Big firms struggle with internal politics and innovation inertia, issues likely to worsen as innovation speeds up. Whether it’s the “Innovators' Dilemma” or the “Law of diminishing returns,” the fact is that big companies have trouble keeping up with the times.
And while big companies have vastly more resources than small startups, startups have almost infinite creativity and agility and can move and adapt quickly. My gut tells me that 18-24 months from now, there will be strong GenAI-native startup companies and products emerging into the spotlight that invoke “aha” moments for consumers and businesses. I also predict many of you will say, “I had that idea!” Well, maybe you should have gone for it.
The resounding message from the panelist was, “Don’t sit on the sidelines!” Seize upon your creative powers and instincts and for change the world. The tectonic shifts that are happening now with GenAI haven’t been seen since previous tech revolutions like Mobile (2010), the Internet (1995), and the personal computer (1984). Now is the time.
Here’s wishing you a happy, creative, and generative 2024.
To learn more about MVL, read our manifesto and let’s build a company together.
We are with our founders from day one, for the long run.
