<?xml version="1.0" encoding="UTF-8"?><rss xmlns:dc="http://purl.org/dc/elements/1.1/" xmlns:content="http://purl.org/rss/1.0/modules/content/" xmlns:atom="http://www.w3.org/2005/Atom" version="2.0" xmlns:cc="http://cyber.law.harvard.edu/rss/creativeCommonsRssModule.html">
    <channel>
        <title><![CDATA[Stories by Jacques Nyilinkindi on Medium]]></title>
        <description><![CDATA[Stories by Jacques Nyilinkindi on Medium]]></description>
        <link>https://medium.com/@j_nkindi?source=rss-fa7aa5fd9fce------2</link>
        <image>
            <url>https://cdn-images-1.medium.com/fit/c/150/150/1*RU3zPbFjyzeRIajcr4A02g.png</url>
            <title>Stories by Jacques Nyilinkindi on Medium</title>
            <link>https://medium.com/@j_nkindi?source=rss-fa7aa5fd9fce------2</link>
        </image>
        <generator>Medium</generator>
        <lastBuildDate>Thu, 19 Nov 2026 06:57:36 GMT</lastBuildDate>
        <atom:link href="https://medium.com/@j_nkindi/feed" rel="self" type="application/rss+xml"/>
        <webMaster><![CDATA[yourfriends@medium.com]]></webMaster>
        <atom:link href="http://medium.superfeedr.com/" rel="hub"/>
        <item>
            <title><![CDATA[The Future of Software Pricing]]></title>
            <link>https://medium.com/@j_nkindi/the-future-of-software-pricing-1a2c37ad3018?source=rss-fa7aa5fd9fce------2</link>
            <guid isPermaLink="false">https://medium.com/p/1a2c37ad3018</guid>
            <category><![CDATA[software-development]]></category>
            <category><![CDATA[ai]]></category>
            <category><![CDATA[pricing-software]]></category>
            <dc:creator><![CDATA[Jacques Nyilinkindi]]></dc:creator>
            <pubDate>Tue, 23 Jun 2026 21:38:36 GMT</pubDate>
            <atom:updated>2026-06-23T21:38:36.846Z</atom:updated>
            <content:encoded><![CDATA[<figure><img alt="" src="https://cdn-images-1.medium.com/max/1024/1*pGcnFgBZ14WiQvBSesLDsQ.png" /></figure><p>A developer asked me a question I couldn’t shake. If AI writes the code faster — if the release that took six months now takes two weeks — shouldn’t the software get cheaper? And if it doesn’t, what exactly was the point?</p><p>It’s the right question. It’s also a trap, because it assumes there’s one answer. There isn’t. There are three, they’re already diverging, and which one you live in depends on decisions being made right now by people who mostly aren’t thinking about it as a decision. So let me lay out the three futures, the forces that select between them, and the bet I think an engineering lead should actually place.</p><h3>First, kill the assumption</h3><p>The intuition is clean: cost of production fell, so price should fall. It feels like physics. It isn’t, for two reasons that every engineer already knows but rarely connects to a pricing model.</p><p>The first is that writing code was never the expensive part. The expensive parts of a software product are the things that happen around the code — the rework, the coordination, the review, the support load, the slow tax of keeping a live system alive. Coding is a slice. So when an assistant makes the <em>writing</em> dramatically faster, it is accelerating one stage of a pipeline whose other stages didn’t move. Some moved the wrong way. The METR randomized controlled trial is the uncomfortable data point here: experienced developers working in codebases they knew were measured <strong>19% slower</strong> with AI tooling, while believing they were 20% faster (<a href="https://arxiv.org/abs/2507.09089">Becker et al., arXiv:2507.09089</a>). Whatever you think the speedup is, the honest version is smaller and lumpier than the demo suggested.</p><p>The second reason is the one with a name. <strong>Jevons paradox</strong>: when you make a resource cheaper to use and demand for it is elastic, total consumption doesn’t fall — it rises to swallow the saving. Coal got more efficient and Britain burned more of it. Cloud compute got cheaper per unit and we provisioned a hundred times more of it. Ryan Spletzer put it well writing about exactly this — he wouldn’t run one API gateway more cheaply, he’d <a href="https://www.spletzer.com/2026/03/jevons-paradox-and-the-future-of-software-engineering/">run a hundred of them for different use cases, produce far more value, and spend roughly the same money</a>. Software demand is about as elastic as demand gets. The backlog of things a business wishes its software did is effectively infinite; the only thing that ever capped it was the cost of building.</p><p>Hold those two facts together — coding is the minority cost, and the saving gets reinvested rather than returned — and the “where’s my discount” question stops being a paradox and becomes a fork. The saving is real. Where it <em>goes</em> is undecided. That’s the whole story.</p><h3>Future One: the price actually drops</h3><p>The optimistic future. Competition forces vendors to pass the saving through, sticker prices fall, and the buyer’s budget shrinks.</p><p>This is the future people instinctively expect and the one I think is least likely for established software — but it isn’t fantasy. It shows up in one specific place: where AI collapses the cost of an <em>alternative</em>. The build-versus-buy maths has genuinely flipped. An internal tool that was a three-month, six-figure project is now a one-to-two-week build. The moment a buyer can credibly threaten to build the thing themselves, the vendor’s pricing power erodes whether or not the vendor wants it to. The discount here doesn’t arrive as a markdown on the invoice you already pay. It arrives as the option to stop paying it.</p><p>The force that selects this future is <strong>substitutability</strong>. Commodity software with no moat gets cheaper because the buyer’s outside option got cheaper. Software with real switching costs, real data gravity, real integration depth does not.</p><h3>Future Two: the price holds, but you get more for it</h3><p>The most probable future, and the one the industry is quietly converging on. Prices stay flat. The saving comes back as faster roadmaps, better reliability, more frequent improvement, more tailored functionality at the same spend.</p><p>This is the future the practitioners building it actually describe. Yoni Michael, CTO at Typedef, told CIO that AI lowering the cost of routine code <a href="https://www.cio.com/article/4151988/ai-is-altering-the-economics-of-software-development-but-who-gets-a-cut.html">should benefit customers, but the most important benefit does not have to be an immediate price cut</a> — it can be faster delivery, better support, more frequent product improvement. Asia Solnyshkina, who runs a development firm, was blunter about why her prices aren’t dropping: code generation was never the main source of value, and her clients are paying for judgment, architecture, integration, and accountability, none of which you solve with a prompt.</p><p>I find this credible because it matches what software actually is. But “you get more for the same money” is only a benefit if the <em>more</em> is something you wanted. More features you don’t use is not value; it’s bloat with a changelog. The version of Future Two worth wanting is the one where the dividend is paid in quality — fewer defects, faster fixes, stronger security — not feature count.</p><p>The force that selects this future is <strong>whether the buyer demands the dividend in a form they can see</strong>. Which is exactly where Future Three comes from.</p><h3>Future Three: the price holds, and the software quietly gets worse</h3><p>The future nobody is selling and everybody should be watching. Prices hold, the vendor pockets the margin, and the buyer inherits a slow erosion of quality they can’t see on the invoice.</p><p>This is the failure mode the security people are already flagging. George Manuelian of RapidFort told CIO that teams are shipping AI-generated code pulling in <a href="https://www.cio.com/article/4151988/ai-is-altering-the-economics-of-software-development-but-who-gets-a-cut.html">outdated or vulnerable open-source libraries — code that works functionally but wouldn’t pass a proper security review</a>, and that the volume of unverified code entering production is rising faster than the processes to validate it. CIO’s own reporting around the same period put a number on the drag from a different angle: a meaningful share of AI productivity gains gets <a href="https://www.cio.com/article/4157471/40-of-ai-productivity-gains-lost-to-rework-for-errors.html">lost to rework fixing errors</a>. In this future the speed is real and the saving is real, but it’s spent generating volume the buyer then has to live with — more surface area, more dependencies, more latent risk — at the same price as before.</p><p>The force that selects this future is <strong>opacity</strong>. Most vendors don’t disclose their AI use in development. Jayanand Sagar of Hyperbola Network put the buyer’s blind spot plainly: that productivity increase <a href="https://www.cio.com/article/4151988/ai-is-altering-the-economics-of-software-development-but-who-gets-a-cut.html">goes straight to margin unless the customer demands something in exchange — and most buyers never ask</a>.</p><h3>Which future you get is a choice, and mostly not yours</h3><p>Here’s the part that matters for anyone leading engineering rather than just doing it. These three futures aren’t a forecast you wait on. They’re a set of outcomes selected by behaviour — by whether buyers ask, whether vendors disclose, whether teams treat quality as the deliverable or treat shipped-volume as the deliverable.</p><p>If you sit on the building side, you decide which future your customers get. The honest move is to spend the saving on the things that don’t demo well — the tests, the review, the dependency hygiene — and let the dividend show up as reliability. That’s Future Two by construction, and it’s also the only durable moat once everyone has the same coding speed.</p><p>If you sit on the buying side, Sagar’s advice is the most actionable thing in any of the source reporting: pull your top vendor contracts, ask for a summary of delivery performance over the past year, and notice the gap. If the release cadence sped up and the invoice didn’t move, that gap is your leverage — and the form to claim it in is faster delivery commitments, tighter SLAs, and security guarantees, not just a markdown.</p><h3>The bet I’d place</h3><p>Long run, I don’t think the price of software falls much. The cost structure won’t allow it and Jevons won’t allow it. The benefit to the buyer is real, but it arrives disguised — as quality you have to demand, and as the leverage to build instead of buy.</p><p>So the budget doesn’t shrink. The <em>value per dollar</em> moves, and it moves toward whoever is paying attention. The buyers who treat AI as a reason to renegotiate for quality will land in Future Two. The ones who assume the discount is automatic will wake up in Future Three, paying yesterday’s price for software that’s quietly carrying more risk than it used to.</p><p>Cheaper was never the promise worth wanting. Better for the same money was, and that one only arrives if someone insists on it.</p><p>— <br>JN</p><img src="https://medium.com/_/stat?event=post.clientViewed&referrerSource=full_rss&postId=1a2c37ad3018" width="1" height="1" alt="">]]></content:encoded>
        </item>
    </channel>
</rss>