Markets remain obsessed with AI, which continues to suck up investor attention and capital like a black hole. It’s easy to dismiss it all as hype — and there’s plenty of that out there — but I’ve found myself increasingly convinced that there really is something very significant underway.<\/p>\n\n\n\n
Not so much because of the undeniable seduction this thing has for one’s inner sci-fi nerd, but more because my hands-on experience with it has given me a clearer and more nuanced perspective. And I’ve been neck-deep in it in recent months with a couple of projects, the main one being a complete rebuild of the Strawman front-end using an AI coding agent.<\/p>\n\n\n\n
Which, as a quick aside, I’m thrilled to report is progressing nicely. Granted, I\u2019ve been running on the classic developer timeline of being permanently “one week away” from launch, though I take some solace in that apparently being a common experience when building software, where the last 5% of the build takes 90% of the time and effort. Moreover, we really are very, very<\/em> close, and I’m committed to having it in your hands before the month is out, come hell or high water!<\/p>\n\n\n\n Anyway, it’s been an extremely enlightening experience and helped evolve my thinking on the whole AI thing, so I thought I’d share a few field notes. <\/p>\n\n\n\n And that, I suppose, is the first thing to say. I think the only way to really understand this thing is to actually get your hands dirty. That requires more than a casual play with a free chatbot. Yeah, it’s intimidating and frustrating, but there is no better way to truly understand exactly what AI can and can’t do.<\/p>\n\n\n\n Even then, it’s been fascinating to experience how fast things are improving. Sure, the tech is far from perfect; it hallucinates, it gets stuck in logic loops, and it makes silly mistakes. But it feels like every other day there’s a patch or an update or a new model, and one by one many of the shortcomings become far less bothersome. <\/p>\n\n\n\n Judging AI by what it cannot do today is a trap; you have to judge it by the trajectory it\u2019s on. And not in some distant, far-off way either. Experiencing the granular, incremental gains and the pace at which they’re rolling out really does behoove us to be less limited in our imaginations.<\/p>\n\n\n\n I do come across plenty of pushback on some of this, but when you dig deeper it seems this most often comes from those who have used a free, low-tier model a year or more ago, given it a lazy one-sentence prompt, and then been disappointed it didn’t return with a perfect response.<\/p>\n\n\n\n Like most things in life, it’s worth paying up for quality if you can. You’ll definitely notice the difference. More importantly, you have to abandon the “one-shot” fallacy. AI isn’t a Google search; it’s a conversation. If you want a great result, you have to invest the time to iterate. Expect failures and restarts, and accept that you need to be the main driver of progress.<\/p>\n\n\n\n It’s not God in a box (yet).<\/p>\n\n\n\n You<\/em> need to act as the lead architect. Especially with software development, you need a clear, high-level mental model of the system you are trying to create so you can direct the AI in a targeted fashion. The best way to treat a modern coding agent is like a brilliant, tireless junior developer. It is spectacular at handling the heavy grunt work, and can even act as an expert specialist for specific tasks, but it lacks macro-judgement. You still need to provide strict context, give explicit instructions, and double-check the work before you deploy it.<\/p>\n\n\n\n Remember, it\u2019s a force multiplier, not a magic trick. AI is just a tool, and like all tools, it amplifies the capability of the person holding it. If you give the exact same model to two different people, you will get two vastly different outcomes. It multiplies your efforts; it doesn’t obviate them.<\/p>\n\n\n\n On a practical level, it helps to understand that, at their core, these models are stateless <\/em>entities. Every single call to a model is a completely fresh start. It has no memory of what you talked about five minutes ago, and any illusion of “memory” only exists because the software layer you are using is quietly injecting your chat history into the prompt behind the scenes. It is essentially a genius suffering from short-term amnesia; brilliant in the moment, but entirely dependent on the context you feed it.<\/p>\n\n\n\n The point here is that I think there is a genuine art in ensuring the context and instructions are fit for purpose. And that’s something you only get a sense of by experimentation and experience. <\/p>\n\n\n\n At first glance, none of this sounds relevant to investing, but I’d argue that a clearer mental map of this technology is a potent way to ground you in reality, and help you navigate all the wild claims and noise you’ll come across.<\/p>\n\n\n\n It’ll also allow you to be more forgiving of what are likely to be rather ugly business fundamentals for companies carving out a niche in this space. Not that you should ignore those things, far from it, but the reality is that almost all of the world’s defining tech giants looked like cash-burning dumpster fires in the early days. When historical financial numbers can’t tell you the whole story, you have to develop an informed view of what the technology actually means. And you’ll never get that just by simply reading analyst reports.<\/p>\n\n\n\n The other angle here is using AI to augment your investment process, which I think is absolutely valid… to a point. I suspect we are about to enter an era where the market is flooded with investors making buy and sell decisions based on shallow, one-sentence prompts given to basic LLMs. They will mistake fast, detailed outputs for deep insights, creating a new brand of confident ignorance.<\/p>\n\n\n\n Again, because AI is a force multiplier, it amplifies the quality of the operator. If you feed it vague context, it will give you confidently average answers. The true investing edge over the next decade won’t belong simply to those who use AI, but to those who know how to drive it intelligently. In the same way that a seasoned developer using AI is an order of magnitude more capable than a novice using the same tools, the same can be said for investors.<\/p>\n\n\n\n My advice is that you have to be highly targeted in your prompting, using a well-curated knowledge base, and remain fiercely disciplined about verifying the output. The tool is revolutionary, but you’re the one that needs to call the shots and make the final judgement.<\/p>\n\n\n\n In the end, AI changes the speed of the game, but it doesn\u2019t change the rules. The technology will keep compounding at hyperspeed, but your competitive edge remains exactly what it has always been: a clear mental model, a healthy dose of scepticism, and the willingness to get your hands dirty.<\/p>\n\n\n\n Strawman is Australia\u2019s premier online investment club.
Members share research & recommendations on ASX-listed stocks by managing Virtual Portfolios and building Company Reports. By ranking content according to performance and community endorsement, Strawman provides accountable and peer-reviewed investment insights.<\/em><\/p>\n\n\n\n