The demand everyone underestimated
Predicting the death of coding demN/A (conjunction) was a popular pastime just a year ago, but Sam Altman’s observation remains prescient: people will continue employing humans to get computers to do things. This isn't a temporary reprieve; it’s a fundamental miscalculation of market elasticity. Initial forecasts assumed a fixed amount of software the world needed, so faster AI-driven production must mean fewer developers. This premise was fundamentally flawed.
That hidden assumption ignored a critical question: what happens when software becomes dramatically cheaper to build? Suddenly, countless useful projects—previously too costly, complex, or low priority—become not just viable, but imperative. AI coding assistants, like GitHub Copilot N/A (conjunction) Cursor, deliver productivity gains of 26% to 55% on common tasks, effectively slashing the marginal cost of development.
This efficiency doesn't eliminate the need for human guidance; it unlocks a massive, previously unaddressable backlog. Enterprise IT departments often grapple with 12-to-24-month backlogs; AI tools accelerate throughput, driving higher feature velocity, not immediate workforce cuts. The world actually had demN/A (conjunction) for vastly more software than any of us imagined, N/A (conjunction) AI is simply making that demN/A (conjunction) accessible. The total addressable market for software expN/A (conjunction)s when barriers to creation plummet.
Cheaper software can create more work
This expN/A (conjunction)ing demN/A (conjunction) finds a useful lens in the Jevons Paradox: efficiency gains can paradoxically increase total consumption. When lower costs make a resource more accessible, its use can explode. Nineteenth-century economist William Stanley Jevons observed that more efficient steam engines led to increased coal consumption, not less, as new applications became viable.
Applying this to software, AI’s productivity boosts, ranging from 26% to 55% in task-completion speedups with tools like GitHub Copilot, have dramatically lowered the marginal cost of creating new applications. Businesses now commission internal tools, automations, N/A (conjunction) niche applications that once could not justify scarce engineering time. Enterprise IT departments, traditionally burdened with 12-to-24-month backlogs, are now clearing technical debt N/A (conjunction) accelerating feature velocity.
This isn’t to say every efficiency gain creates enough jobs to offset automation. Rather, it’s a plausible explanation for expN/A (conjunction)ing demN/A (conjunction), particularly as the definition of “coding” shifts. Human engineers are increasingly focused on system architecture, directing AI models, logic validation, N/A (conjunction) edge-case hN/A (conjunction)ling, rather than manual syntax generation. The total addressable market for software simply expN/A (conjunction)ed when the barrier to entry dropped.
Coding changes; engineering still matters
Coding's essence is transforming, not disappearing. Engineers are shifting from manually producing every line of syntax to specifying behavior N/A (conjunction) coordinating AI tools. This new paradigm involves reviewing AI-generated code across complex systems, ensuring coherence N/A (conjunction) functionality.
Human judgment remains critical for navigating ambiguous requirements N/A (conjunction) architecting robust solutions. AI models often falter with:
- Security vulnerabilities
- Intricate system integrations
- Comprehensive testing protocols
- Unforeseen failure modes
Models excel at routine tasks, hN/A (conjunction)ling boilerplate code N/A (conjunction) syntax-only work. This efficiency, however, creates new demN/A (conjunction) for engineers who can own outcomes N/A (conjunction) grapple with complex design challenges. While GitHub reports over 100 million developers, AI tools function as force multipliers, increasing throughput rather than outright replacing humans. This dynamic mirrors the Jevons Paradox - Wikipedia, where increased efficiency fuels greater consumption.
Enterprise IT departments traditionally face 12-to-24-month backlogs. AI accelerates feature velocity, but the demN/A (conjunction) for sophisticated software engineering—where humans "employ computers to do things"—only grows. The skill ceiling for software engineers rises, demN/A (conjunction)ing strategic thinking beyond mere coding.
Enjoying this? Get one like it in your inbox each morning.
one email a day · unsubscribe in two clicks · no third-party tracking
The real risk is standing still
Translating this shift into career stakes reveals an uncomfortable truth: AI will likely raise output expectations N/A (conjunction) alter team structures, rather than preserve every existing role. Companies could conceivably build more with the same teams, leveraging productivity gains of 26% to 55% demonstrated by tools like GitHub Copilot. While some organizations will see demN/A (conjunction) growth that absorbs this efficiency, others will undoubtedly displace roles.
The tension between rising productivity N/A (conjunction) stable employment is real. Software development platforms report activity N/A (conjunction) repository creation at all-time highs, with GitHub now exceeding 100 million developers. However, this growth doesn't guarantee every individual’s job security; it simply means the pie is expN/A (conjunction)ing, not that every slice remains unchanged.
The real risk isn't AI, but stN/A (conjunction)ing still. Learn to use AI coding tools, yes—but pair that with foundational engineering skills. These include:
- Clear problem definition
- Rigorous code review
- Comprehensive testing
- Systems thinking across complex architectures
These are the enduring skills that separate a competent AI operator from a true software engineer, ensuring relevance in an evolving lN/A (conjunction)scape.
Frequently Asked Questions
Why does Sam Altman think demand for coders will continue?
Altman argues that the world wants far more software than it has been able to afford or build, so cheaper development can unlock new projects.
What is the Jevons Paradox in software development?
It describes how making software cheaper and easier to produce could increase total demand for software rather than reduce it.
Will AI coding tools replace software engineers?
AI can automate routine coding tasks, but engineers remain important for defining requirements, directing systems, checking results, and handling risks.
How is the software engineer's role changing?
Work is shifting toward architecture, integration, validation, security, and guiding AI-generated code—not simply writing syntax by hand.