My timeline is full of GPT influencers, so I'd like to push back a bit: LLMs are nowhere as advanced as most people would have you believe. It will be a while before you can automate away most jobs. Why? There are plenty of single-step inferences on the internet: if you want to travel, search a travel site. If you need to do math, find a calculator. But for long chains of long-tail problem solving, this is not true. If you let an agent go away for several cycles it will start to compound its own mistakes and most likely fail to accomplish what you wanted. Or worse, it will make a mess that you'll have to undo. This problem cannot be solved with more computation and more data in an economic way. Not yet, at least. Many startups will find out the hard way this year and next. LLMs have their place in the world of business, just like calculators and spreadsheets. But they won't replace your accountant or your doctor or our lawyer any time soon. They will help larger companies (the Fords and Pepsis of the world) reduce inefficiencies and perhaps eliminate some jobs that are already borderline useful, but it's not a step function. It's the continuation of a trend.