[{"data":1,"prerenderedAt":148},["ShallowReactive",2],{"blog-post-en-we-automated-the-jobs-we-worry-about":3},{"id":4,"title":5,"author":6,"body":7,"category":131,"date":132,"description":13,"excerpt":133,"extension":134,"gradient":135,"icon":136,"meta":137,"navigation":138,"path":139,"seo":140,"stem":141,"tags":142,"__hash__":147},"posts_en\u002Fblog\u002Fwe-automated-the-jobs-we-worry-about.md","We Automated the Jobs We Worry About","mercedes",{"type":8,"value":9,"toc":122},"minimark",[10,14,17,22,31,34,38,41,49,52,56,59,62,65,69,72,91,99,103,106,109,112],[11,12,13],"p",{},"Most writing about AI and jobs is written from a safe distance — by people whose own headcount is not the thing being automated. We don't have that distance. This company exists because a team this small can now operate what used to take a department, and we would rather say that plainly than pretend our optimism is disinterested. The honest version of our position starts with an admission: we are not observing the shift in work. We are an instance of it.",[11,15,16],{},"That is uncomfortable to write, and the discomfort is the point. A company whose entire edge is doing more with fewer people has every incentive to narrate the jobs question into something softer. We'd rather not.",[18,19,21],"h2",{"id":20},"the-claim-we-actually-make","The claim we actually make",[11,23,24,25,30],{},"The argument we ",[26,27,29],"a",{"href":28},"\u002Fblog\u002Fwhy-the-200-dollar-seat-exists","have made before"," is that a small team can now build and run a product that used to require forty people. That is straightforwardly true and we stand by it. What it means for the people who were the other thirty-six is a separate question, and it does not get easier by being left implicit.",[11,32,33],{},"So here is the specific claim, stripped of comfort: the cost of producing competent, conventional software fell by a large factor, and that category of work used to employ a lot of people. Not because any one task was hard, but because the volume was high and the volume was the job. When the volume stops being the constraint, the roles organized around that volume stop being structurally necessary. That is not a prediction. It is already the arithmetic we run on.",[18,35,37],{"id":36},"the-optimistic-half-is-real-and-incomplete","The optimistic half is real — and incomplete",[11,39,40],{},"The reason this is not a doom post is that the same force cuts the other way, and just as hard.",[11,42,43,44,48],{},"The barrier that used to keep most people out of building things — capital, headcount, years of specialized labor — is mostly gone. It has never been cheaper or faster to turn an idea into something that exists and serves real users. We are proof of that. The company making the uncomfortable argument above was made ",[45,46,47],"em",{},"possible"," by the exact technology it is being careful about. Both things are true at once, and anyone selling you only one of them is selling you something.",[11,50,51],{},"Here is where I want to be honest rather than tidy. The optimistic story has a quiet assumption baked into it: that the person displaced from a role can simply go build their own thing. Become the small team instead of the department. And maybe — maybe — that is where this goes. But \"everyone becomes an entrepreneur\" is an easy sentence to write and a hard life to hand someone who did not ask for it. Not everyone wants to found a company. Not everyone should have to. I don't know whether the future that asks this of people is a fair one, and I am suspicious of anyone who claims they do.",[18,53,55],{"id":54},"why-a-fast-moving-company-wants-rules","Why a fast-moving company wants rules",[11,57,58],{},"The instinct in our part of the industry is to treat regulation as friction — a tax imposed by people who don't understand the technology. I think that instinct is both self-serving and wrong, and it's worth saying so from inside a company built on moving fast.",[11,60,61],{},"Guardrails, including real regulation, are not about slowing the technology for its own sake. They are about the gap between how fast capability is moving and how fast institutions, safety nets, and people can adapt. When that gap gets too wide, the cost of it does not disappear — it just lands on whoever is least able to absorb it. A company that profits from the speed is precisely the kind of company that should be willing to say the speed needs managing. The people closest to how fast this moves have the least excuse to stay quiet about it.",[11,63,64],{},"That is not a regulatory proposal; I am not pretending to have drafted the policy. It is a disposition. When the choice is between \"let it run and sort out the damage later\" and \"build in time for people to adapt,\" we come down on the side of adaptation, even when it is against our own short-term interest to say so.",[18,66,68],{"id":67},"what-makes-this-more-than-a-blog-post","What makes this more than a blog post",[11,70,71],{},"Positions are cheap. The reason to trust one is whether it costs the person holding it anything, and whether it shows up in behavior.",[11,73,74,75,79,80,83,84,90],{},"Ours shows up in how we build. AI drafts; humans approve; a red pipeline stops a release no matter how confident the model was, and the most impactful architectural decisions are made by a person — the discipline we argued for in ",[26,76,78],{"href":77},"\u002Fblog\u002Fthe-trust-problem-coming-for-vibe-coded-software","the trust post"," is the same discipline that keeps \"move fast\" from becoming \"move carelessly.\" It also shows up in a line we will draw clearly: we use AI to automate the ",[45,81,82],{},"building"," of software, not to make decisions about people. Our products do not score, rank, or profile individuals. Kleppmann and Riccomini call the alternative ",[26,85,89],{"href":86,"rel":87},"https:\u002F\u002Fwww.oreilly.com\u002Flibrary\u002Fview\u002Fdesigning-data-intensive-applications\u002F9781098119058\u002F",[88],"nofollow","\"algorithmic prison\""," — automated judgment that quietly excludes people from jobs, credit, and services with little recourse — and we want no part of building it. On the same logic, we collect the minimum data we need and do not sell it, because collected data is a liability before it is an asset: a standing risk to the people it describes, not just a resource for us. And because we build with small teams, we treat the jobs question as ours to sit with — in who we hire, how we grow, and what we are willing to automate — rather than as an externality for someone else's quarterly report.",[11,92,93,94,98],{},"We've also stopped keeping this as an internal value and written it down as a standing position, in plainer language than this post, on our ",[26,95,97],{"href":96},"\u002Fresponsible-ai","Responsible AI page",". Not because a page fixes anything, but because a position you'll put your name to in public is harder to quietly abandon than one that only lives in a culture deck.",[18,100,102],{"id":101},"the-part-we-wont-resolve","The part we won't resolve",[11,104,105],{},"I could end this with a reassuring line about how it all works out. I'm not going to, because I don't believe the people writing those lines know either.",[11,107,108],{},"We are optimistic about where this goes — genuinely, and from direct experience of the doors it opens. We also refuse to be careless about how it gets there, because we have watched, from the inside, how quickly the ground moves. If those two things sound unresolved held together, that's because they are. We would rather hold the tension honestly than collapse it into a slogan that makes us sound better than we've earned.",[110,111],"hr",{},[11,113,114],{},[45,115,116,117,121],{},"Reference: Martin Kleppmann & Chris Riccomini, ",[26,118,120],{"href":86,"rel":119},[88],"Designing Data-Intensive Applications, 2nd edition"," (O'Reilly, 2026), ch. 14, \"Doing the Right Thing.\" The \"algorithmic prison\" framing and the view of data as a liability rather than an asset are theirs; the stance we take on them is our own.",{"title":123,"searchDepth":124,"depth":124,"links":125},"",2,[126,127,128,129,130],{"id":20,"depth":124,"text":21},{"id":36,"depth":124,"text":37},{"id":54,"depth":124,"text":55},{"id":67,"depth":124,"text":68},{"id":101,"depth":124,"text":102},"business","2026-10-12",null,"md","linear-gradient(135deg, rgba(79,70,229,0.4), rgba(99,102,241,0.2))","i-heroicons-scale",{},true,"\u002Fblog\u002Fwe-automated-the-jobs-we-worry-about",{"title":5,"description":13},"blog\u002Fwe-automated-the-jobs-we-worry-about",[143,144,145,146],"AI","Future of Work","Automation","Responsible AI","ckJru3nu9ipkm-mBGcHKdmZWjH2nAGQf-aM0uAVdJkA",1792082304686]