TL;DR
Get kitchen staples and gadgets delivered free with Prime
- Fast, free delivery on millions of items
- Prime Video, Amazon Music and more included
- Member-only deals all year
Sanket Saurav’s essay argues that AI coding tools have made software development cheaper while making AI-powered products more costly to run. He says companies are still copying familiar software categories rather than redesigning products around the new economics; the supplied report ends mid-sentence before completing its proposed remedies.
Sanket Saurav’s essay, “The End of Rationing,” argues that AI coding tools have lowered the cost of building software, but that product design across the industry has not yet caught up. The piece matters as a critique of the current wave of AI products: Saurav says many add AI to familiar categories instead of rethinking what software could be when features are cheaper to build.
Saurav describes a shift in software economics around the arrival of tools such as Claude Code. Before that shift, he argues, engineering time was scarce and expensive, while the cost of running most software was comparatively low. Companies rationed development effort through practices such as the minimum viable product and “do one thing” product strategies. He says those constraints helped shape products and categories that later competitors copied.
In the newer model Saurav describes, coding agents and AI tools reduce the labor needed to build features, while products that depend on large language models incur ongoing model costs for each user or task. He characterizes this as a move from high research-and-development costs and low marginal operating costs toward leaner development and more substantial per-use costs. This is his analysis of the economics, not a quantified industry-wide measurement in the supplied material.
The essay points to GitHub pull requests and the familiar CRM structure associated with Salesforce as examples of patterns that became standard. Saurav says current AI companies often reproduce existing products or categories, including vertical versions of agent software, rather than starting from first principles. The supplied excerpt ends after the words “The second is less obvious: pack,” so it does not include the complete account of the two forms of change he says are needed.
Why AI Products Still Look Familiar
The argument challenges the assumption that cheaper software development will automatically produce better or more varied products. If Saurav is right, the main obstacle is no longer just the cost of writing code: it is the way companies organize product work, sell new categories, and decide what to build. That distinction matters to software founders and investors deciding whether to improve an established product or invent a different one.
There are practical limits to the essay’s premise. AI features carry per-user operating costs, software still needs maintenance, and newcomers face established competitors with customer relationships and distribution. Saurav acknowledges these constraints. His central point is narrower: they do not fully explain why many new products retain familiar designs and workflows.
The piece offers a framework for evaluating AI product claims: ask whether a product changes the underlying workflow or merely adds a conversational interface to an existing one. That is a useful question for buyers as well as builders, though the essay does not provide comparative performance data to establish that a redesigned product will be better.
As an affiliate, we earn on qualifying purchases.
From Cheap Software to Costly Inference
Saurav frames the change as a reversal of the economics that shaped conventional software. In the earlier model he describes, engineering teams were a major fixed expense, but serving another user could be relatively cheap. AI products alter that calculation because running models can add variable costs as usage grows. The supplied source gives no general cost figures or comparison period for this shift.
To explain why product practices may lag behind new technology, Saurav compares software with the adoption of electricity in US manufacturing. He says factories initially replaced steam power while keeping existing layouts, and that larger productivity gains came when factories reorganized around smaller electric motors at individual machines. He uses the analogy to argue that software companies may need to redesign their development and product systems, not just insert AI into existing ones.
The essay also identifies organizational and commercial barriers: procurement teams tend to buy established categories, investors often back familiar “X for Y” business models, and large companies may be structured around existing headcount and processes. These are Saurav’s explanations for slow change, not independently documented findings in the material provided.
““The short answer is rationing. It’s over, and we haven’t caught up.””
— Sanket Saurav, in “The End of Rationing”
software development automation software
As an affiliate, we earn on qualifying purchases.
As an affiliate, we earn on qualifying purchases.
Limits of the Essay’s Case
The supplied report is an essay, not a market study, and it does not provide figures showing how much AI has reduced development costs or how operating costs compare across products. Its statements about industry practice and investor preferences are Saurav’s assessments; the material does not include responses from companies, investors, or procurement teams.
The source excerpt is also incomplete. It introduces two forms of change, explains the first through examples involving code collaboration and CRM, and then cuts off at “The second is less obvious: pack”. The second proposal and the essay’s ending cannot be reported from the supplied text. The source material does not specify when the essay was published, and it does not establish whether the products Saurav describes have since changed.
As an affiliate, we earn on qualifying purchases.
The Unfinished Second Proposal
The next step for readers seeking the full argument is to consult Saurav’s original essay, which may contain the missing continuation and closing section. Based on the supplied excerpt alone, no product launch, company decision, or scheduled industry milestone follows from the article.
The broader question raised by the essay is whether software companies will build products around new human-and-agent workflows, rather than simply add AI to existing categories. Evidence would include products that materially change how users work, alongside clear disclosure of their operating costs and maintenance needs. The excerpt does not name a particular company expected to make that shift.
As an affiliate, we earn on qualifying purchases.
Key Questions
What does “the end of rationing” mean in the essay?
Saurav uses the phrase for a shift in which AI coding tools make it cheaper to build software features, reducing the need to ration scarce engineering time. He argues that companies have not yet changed product design to match.
Does the essay say AI software is cheaper to operate?
No. Saurav argues that building software has become less costly while running products that use large language models can carry meaningful costs per user or task. The supplied material does not give a universal cost estimate.
Which established products does Saurav discuss?
He cites GitHub’s pull-request workflow and Salesforce’s lead, opportunity, and pipeline model as examples of established patterns that competitors adopted. He argues that future products could rethink those categories rather than simply add AI to them.
What is the essay’s second proposed form of change?
The provided excerpt does not say. It ends just as Saurav begins that point, with the incomplete phrase “The second is less obvious: pack”.
Is the essay’s account of the industry independently verified here?
No. The supplied source presents Saurav’s analysis and examples but includes no independent data, company responses, or industry-wide cost comparison. Its claims should be read as the author’s argument.
Source: hn
Evergreen bestsellers Picks
bestsellers
As an affiliate, we earn on qualifying purchases.
