1984 Ventures

Dear investors, founders and friends

Vladimir Lenin famously said, "There are decades when nothing happens; and there are weeks when decades happen." 2024 was definitely a year when decades of technological progress in AI happened. What started as experimental applications of AI quickly evolved into practical solutions delivering huge ROI and fundamentally changing various industries. The revenue growth rates of these companies have been astounding, their business models are subversive, and the breadth of sectors being transformed is limitless. 2024 was the year when the magnitude of the AI opportunity became evident to all.

The Challenge for Early Stage Startups

The paradox of the moment is that while opportunities are boundless, so is competition. In 2024, we saw many more startups pursuing each idea than years past. Additionally, the tech incumbents are clear-eyed about the opportunity and are much better at executing than their counterparts in the past. To win in this environment, startups need a few critical elements:

• Laser focus on customer needs: If the experience doesn't feel like magic providing 10x ROI, it's unlikely to sell.
• Build more than just AI: The best problems to go after are the ones where there's a major distance between the customer and the underlying models. "Fine tuning" a model isn't a moat--particularly as models improve to be good enough for most tasks on their own. Instead, think of custom integrations, complex workflows, and layers of UI.
• Agile technical capabilities: Speed of shipping and adapting to new frameworks has never been more critical
• White Ocean: Competing with incumbents is going to be a challenge (unless there is an innovator’s dilemma). Find a vertical that a few other players understand.
• A Travis Kalanick-like attitude of relentless determination and drive to dominate at all cost

The Technical Mess

With the technology landscape evolving at breakneck speed, things are far from stable. For example, this year Anthropic's Claude surpassed OpenAI in development capabilities, something that was unthinkable in 2023. (In 2025, watch out for the Chinese models Qwen and DeepSeek gaining momentum). Additionally, the current state of AI tooling remains remarkably primitive, reminiscent of the early internet when basic HTML editors and FTP clients were considered cutting-edge development tools. We first invested in devtools with Posthog, one of the most widely used companies in our portfolio. We are excited to continue pursuing this investment category in Fund 3 which started with Cline, currently working out of our office in San Francisco and giving Cursor a run for its money as the best performing coding assistant.

Late Stage Startups Must Adapt

With such a seismic shift in tech capabilities and expectations, late-stage startups must adapt. AI presents these companies with three challenges: their offerings are quickly becoming outdated, their products are much easier to build today (thanks to autonomous coding), and their markets are often becoming commoditized. We have seen some portfolio companies suffer as a result. But we have also seen winners in the portfolio adapt and thrive.

The Challenge for VCs

The same challenges apply for Venture Capitalists. Expertise in software investing today mirrors expertise in Semi-conductor investing in the late 90s—it will soon become obsolete. To stay competitive in this new world order, VCs need a technical understanding of AI to ascertain where the puck is heading. And with the breadth of opportunities to consider, success now demands specialized sector focus. Last, with so much consolidation and fundraising happening at later stage, it takes discipline to remain small, nimble and stage focused on pre-seed. Yet herein lies our opportunity. As Eric Vishria of Benchmark recently quipped on why stay nimble in Venture Capital: “Because when it works, it’s really magical.”

An Inevitable Correction

The glut of financing won't last forever. Eventually, investors pumping billions into AI models will realize they're repeating the telecom overbuild of the late 1990s. And given the personalities involved, drama will likely be spectacular—grab your popcorn as the xAI/OpenAI drama unfolds like a real-life "Silicon Valley." Yet don't expect us to stop investing when the bubble bursts: the hurdle to AI adoption in the enterprise is less about technology breakthroughs and more about friction and incentives which will take time to sort out, and the moment when investors turn skittish is often when new technology transforms from hype to reality.