LLMs will demolish barriers for software creation and unleash the next great technology cycle, marking a kind of Gutenberg moment for the software industry
scholarly, creative, economic, etc—are going to fall away, as people are freed to do things only limited by their imagination, or, more practically, by the old costs of producing software” https://twitter.com/... @om : @theoasisAI @Spotify All the hype around AI and ChatGPT needs some context. Paul Kedrosky and Eric Norlin in their essay “Society's Technical Debt and Software's Gutenberg Moment” weave a compelling link between software, economics, and how humans value software. https://skventures.substack.com/ ... https://twitter.com/... Paul Kedrosky / @pkedrosky : And here is the essay again. https://skventures.substack.com/ ... And a related followup. https://skventures.substack.com/ ... Steve O'Grady / @sogrady : in addition to yegge, btw, you should be reading @defrag and @pkedrosky's, ah, less optimistic take on these technologies https://skventures.substack.com/ ... Chris Anderson / @chr1sa : Both agree that call centers and research analysts will disrupted by AI, and that teachers and psychiatrists are probably safe. Lee also attempts to identify which professions will be hugely enhanced by AI. Anyway, the SK essay is very interesting https://skventures.substack.com/ ... Brink Lindsey / @lindsey_brink : Among the most interesting pieces on the implications of large language models I've read yet: the coming collapse in the cost of producing software. https://skventures.substack.com/ ... Paul Kedrosky / @pkedrosky : Good comments, both direct and parallel, to our recent note on the coming LLM-induced changes in software (https://skventures.substack.com/ ...) https://twitter.com/... @policytensor : This is the first smart thing I've read on AI. Bullish, but in a serious way. https://skventures.substack.com/ ...
Context & Ripple Effects
Paul Kedrosky and Eric Norlin's essay 'Society's Technical Debt and Software's Gutenberg Moment' lands in the middle of a March 2023 argument about what LLMs actually are. A day earlier, Sourcegraph's blog made the same bet from the practitioner side, calling LLMs the biggest thing since IDEs and Stack Overflow rather than a crypto-style fad — the two pieces together frame software creation itself as the thing being repriced.
The claim has a visible skeptic: Emily M. Bender, whose co-authored paper on what LLMs can and cannot do anchors the linguistic critique of treating these systems as reasoning engines. The Gutenberg framing matters because it stakes out an economic position — collapsing the old costs of producing software — rather than a capability position, which is exactly the axis on which that debate runs.
First-order effects
- The immediate audience is anyone whose software demand was priced out by development costs: the essay argues people become limited by imagination rather than by the old economics of producing software, shifting who can commission working tools from companies with engineering teams to individuals.
Second-order effects
- If software supply cheapens this way, incumbents whose pricing rests on scarce development capacity face pressure to compete on distribution and data instead of build cost — the dynamic Karpathy later sharpened by arguing LLMs disproportionately benefit regular people, with a muted, slower impact on corporations and governments.
Third-order effects
- If the pattern holds, software stops behaving like a manufactured good and starts behaving like content — abundant, personalized, and cheap — which would push value toward whoever controls models, inference, and trust rather than whoever writes code, and force policy debates like the one over AI-as-normal technology versus humanlike intelligence to treat democratized software creation as infrastructure rather than novelty.
The trend: Software creation is entering a cycle where LLMs collapse production costs the way printing collapsed reproduction costs, turning scarcity of builders into abundance of built things.