I keep seeing the phrase “meat proxy” show up in my feed, usually from an engineer joking that their job now is to sit between a model and a keyboard and click accept. It gets a laugh because there is truth in it. A good friend of mine, Suresh Kakarla, wrote recently about how this is reshaping engineering organizations, with teams getting smaller and more senior as AI takes on more of the execution, and a real risk that companies stop developing the next generation of engineers because junior work looks redundant now. I think he is right about the organization. What has been sitting with me since is the same question at the level of one person’s thinking, not a team’s structure.
Twenty-plus years in enterprise technology taught me that judgment is not something you are handed. It is built the slow way, by writing the wrong query, watching it fail in a way you did not expect, and sitting with that failure long enough to understand why. Plenty of that friction was never fun and a lot of it was avoidable in hindsight, but it is also where the instinct came from that let me later look at a system and know, before I could fully explain it, that something was off. AI is very good at removing that friction. It is much less good at telling you when removing it costs you something.
The comparison people reach for is the calculator, and it is a fair one as far as it goes. Nobody mourns doing long division by hand. But a calculator does not decide what problem you are solving, or whether the number it gives you is the right number to be asking for in the first place. A model that writes the function, picks the approach, and explains its own reasoning back to you is doing something closer to thinking on your behalf than computing on your behalf, and the difference matters more the earlier it happens in someone’s career.
I use AI heavily myself and I am not writing this as an argument against it. I believe strongly that the direction of travel, open models, open weights, tools that reason alongside you, is good for the industry and good for the people in it. But I have started noticing my own habits around it. I try to write down what I think the answer is, or at least the shape of it, before I ask a model for one, and I treat what comes back as a second opinion to argue with rather than a first draft to accept. Some days I skip that discipline because I am tired or the deadline is close, and I can feel the difference in how much of the problem I actually understood afterward.
None of this is really about the tools. It is about whether the habit of thinking something through gets practiced or gets skipped, and a model that is fast, confident, and usually right makes skipping it the path of least resistance. That is the part the “meat proxy” joke gets right, whether or not the people making it mean it that seriously. The part it misses is that this is a choice being made one small moment at a time, by individuals, and it is still ours to make differently.
I would rather stay the kind of engineer, and now advisor, who can tell you why an answer is right, not just that a model produced it. I suspect a lot of people reading this feel the same pull and would be glad to compare notes on how they are handling it.