← Essays

Machines of Loving Grace, Two Years On

The Provider's Question

Last in a series reading Dario Amodei's "Machines of Loving Grace" from 2026. The earlier essays were about value functions, datasets, and gates. This one is about the final section, work and meaning, and it is the one I have the most standing to write and the least distance from. The essay's author admits this section is the hardest and offers the fewest answers. He's right. Here is what I think he missed anyway.

I should say who is talking, because for this topic it matters. I am the son of two factory workers, the first in my family to go to college. I have spent 25 years in knowledge work that my parents' world would have called a desk miracle, and I know exactly what I escaped, because the fear of falling back is real and it does not expire. When I read a section about what humans will do for meaning when work is optional, I read it from both sides of that line.

A rich person's dilemma

And that is my core critique: the section never once views the world from the everyman's perspective. Contemplating meaning-without-work is only possible from a vantage where work is optional. For most people it is not, and it never has been. It is a rich person's dilemma, honestly arrived at, but a rich person's dilemma all the same. Everyday people work to live and to provide. The pride they take is the pride of making the best of necessity. Ask the people I grew up around where meaning comes from and the answer is not "my tasks were intellectually stimulating." The answer is: I provided for the people I love.

The wrong location

The essay asks the craftsman's question: what happens to meaning when AI can do your tasks better than you can? It is a real question and I feel it personally; I have watched systems I built do things I used to be proud of doing by hand. But the craftsman's question is the wrong location for the wound. For the everyman, meaning was never stored in the tasks. It is stored in the role the tasks fund. Provider. The one who kept the lights on, who got the kids through school, who was the reason the family made it. AI taking the work is therefore worse than the essay imagines, not better: it does not threaten how people spend their days, it threatens who they get to be for their families. The wound is not vocational. It is identity through provision.

Let me put the same point in a cheaper, truer way. People don't fear AI taking their meaning. They fear it taking their advantage.

Positional goods

That word, advantage, is doing real work, and it points at the thing polite futurism won't say out loud. A decent chunk of daily contentment is positional. It is knowing, quietly, that you are better off than you were, and yes, better off than some others. Economists call these positional goods, and their defining property is that they cannot be given to everyone, because they are made of comparison. This is why the standard abundance answer leaves me cold. Here is my own sentence on it, and I will stand behind it: when everyone has the same things, no one has anything, except for the rich, who truly have everything. Universal provision flattens the middle's hard-won distinctions while leaving the top's untouched. That is the asymmetry the UBI debate keeps politely stepping around, and no amount of abundance dissolves it, because the good in question is relative position itself.

There is a structural point about meaning hiding here too. Momentary excitement requires ordinary days. A vacation is a vacation because of the fifty weeks around it; live on vacation and it is just your life, with the same dishes to do. Meaning is built out of contrast, and a world engineered to remove all the contrast, all the necessity, all the stakes, is not obviously a world with more meaning in it. Possibly the opposite.

The vineyard

Since the essay's vision is essentially the Star Trek future, let me grade it on Star Trek's own evidence, because I know that universe well and it agrees with me whenever it looks closely. The franchise's showcase of post-scarcity meaning is Picard retiring to the family vineyard. Look at the vineyard: it is inherited landed property. The best image of meaning that the quintessential limitless-resources universe could produce is aristocracy, position that provision cannot buy you because it comes from lineage. And the canon knows it: there is a scene where a former colleague, living in a trailer in the desert, bitterly contrasts her situation with Picard's fine chateau. Inequality inside paradise, on screen, in the flagship utopia. Meanwhile, notice where every character with abundant meaning actually lives: Starfleet, an institution that manufactures rank, mission, scarcity, and stakes on purpose, inside a society that abolished them. The ship gives everyone a vineyard. And Deep Space Nine, the franchise's conscience, put the whole critique in nine words when Sisko observed that "it's easy to be a saint in paradise." Paradise is where the hard questions about human nature go to hide.

The inventory

I want to be fair about what I am not saying. I am not saying meaning was never in the work; mine partly is, and I am lucky that way. I am saying age taught me the inventory. I used to think I was my title. I am not. The title is what I do. What I am, it turns out, is a husband and a father, and the work is what buys me the resources to be those things well. That reordering took me decades and it is the most valuable thing I know about this topic. It is also exactly the reordering the essay never performs: it audits the tasks and never audits the roles the tasks pay for.

The wrong value function

So here is where the series lands. Four essays ago I said compute multiplies what the data and the value function allow. The work-and-meaning section is the final case of it, and the most human one: the essay is optimizing meaning with the wrong value function. It scores tasks, stimulation, contribution. The people with the most to lose score something else entirely: whether they can still be the provider, still hold their position, still matter to the people they matter to. If powerful AI arrives holding the first scorecard while the world runs on the second, we will solve the wrong problem beautifully.

The open question, then, the one I'd want every builder of these systems to sit with, is not "what will people do all day." It is this: can a society give everyone the standing that providing used to buy, when standing is made of comparison, or does the end of necessity quietly demote the very people who had finally clawed their way to enough? I have watched one family make that climb. I am not willing to wave my hands at what happens to the ones mid-ladder when the ladder gets automated.

That's the series. The essay I've been responding to is ambitious, humane, and wrong in interesting places, which is the highest compliment I give a document. Two years into its test window, my scorecard reads: biology, promising and unresolved. Economics, diverging first. Governance, the right fears and a contradictory strategy. Meaning, the right question asked in the wrong location. I'll keep grading as reality keeps arriving.

Notes and further reading

Jason Stiles has spent 25 years building conversational applications in production, from IVR to agentic AI. He writes a public build log at stiles.one/hedge/build.