The End of Rationing

An essay by Sanket Saurav, published on 13 October 2026, asking why software hasn’t got much better now that a strong coding model costs a couple of hundred dollars a month. His answer is rationing. For decades software was expensive to build and cheap to run, so engineering time was the scarce thing, and the industry’s doctrines grew out of that scarcity: the minimum viable product, the lean startup, do one thing well. The shapes it settled on outlived their reasons. GitHub’s pull request was built for open-source strangers in 2008 and copied everywhere; Salesforce’s leads, opportunities and pipeline became what every CRM looks like.

Now the economics have flipped. Building is cheap, and running is expensive, because every user brings a real model bill, so software starts to look more like manufacturing than SaaS. The result so far is “more software, not different software”: clones of the same agent, a flood of Show HN posts, products from 2015 with a chat box bolted on, and AI products running gross margins well below the old SaaS norm. He blames old habits, buyers who purchase by category, investors who fund “X for Y”, and incumbents who ship their org charts. Some constraints do remain: distribution, trust, maintenance at scale, and the per-user token bill.

His analogy is electrification. American factories took about four decades to get the productivity out of electricity, because at first they swapped one steam engine for one big motor; the gains came when the floor was redesigned around a small motor at every machine. He sees two ways to do the software equivalent: rethink a category from first principles (code review for a world where people read less code, a CRM without the old object model), or package many products together and own the customer, as aggregators do. The rationing ended, he concludes, before the habits did, and the advantage goes to people who never learned to ration.