The AI world is active with two new startups pushing what intelligence means. Recursive released Flint, an AI model that largely built itself, coding on a laptop. Meanwhile, Electric Plant says it can translate what plants 'say' using electricity. These are impressive technical feats, but for investors, the immediate question is simple: who makes money, and how much?
Recursive's Flint is a laptop-sized coding model, a kind of mini-ChatGPT for software development. The significant news is that its own AI research system largely built it. This moves toward AI improving itself, a long-held goal. Flint is free, an early test, which means no revenue yet. It shows AI's potential to automate more complex tasks, but it also suggests a future where other AIs might be the 'creators' of AI.
Then there's Electric Plant, addressing a problem few thought existed. They built an AI-powered translator to decipher the subtle electrical signals plants emit. The idea is that plants 'speak' through these electrical currents, and AI can help humans understand. This is pure research right now, a scientific curiosity. The market for 'plant communication translators' is new. It is a long way from a lab experiment to a product with paying customers.
These projects are at the frontier, exploring the outer limits of AI capabilities. On one side, you have AI building AI, pushing automation and self-improvement. On the other, you have AI trying to unlock nature's secrets, venturing into biology and communication. Both are technically impressive, showing the speed of development in the AI space. But neither offers a clear, direct path to significant revenue or profit in the short term.
This is where we need to follow the money. The real cash in AI right now is not in these experimental applications. It is in the underlying infrastructure: the chips from Nvidia, the massive data centers from AWS, Microsoft Azure, and Google Cloud, and the energy needed to power them. These are the companies selling the shovels in a gold rush, while many of the AI application startups are still prospecting.
For investors, this means differentiating between groundbreaking research and viable business models. Recursive's Flint might eventually lead to tools that cut software development costs for large enterprises, but that is a long road. Electric Plant's technology is even further from commercialization; its applications might be in agriculture or environmental monitoring, but the market size and willingness to pay are unknown. The venture capital pouring into these areas is speculative, betting on long-term potential rather than immediate returns. As we say in Brazil, 'dinheiro na mão é vendaval' (money in hand is like a whirlwind), it can disappear fast if not invested wisely.
The winners today are the companies providing the raw compute power and the fundamental large language models (LLMs, the tech behind ChatGPT). The losers are those who invest heavily in niche, unproven AI applications without a clear path to market or a defined customer base. The valuations of these cutting-edge startups often reflect future speculation rather than current financial performance. The challenge is converting impressive AI demonstrations into actual products that solve real problems for paying customers, at scale.
