Elon Musk on Rogan: “AI will take the computer jobs.” What that means for web developers in 2026
Elon Musk just told Joe Rogan that anything digital — “someone at a computer doing something” — will be taken over by AI. Physical, atom-moving work (plumbing, welding, cooking) lasts longer; computer work is first in line. The episode (#2404) dropped on October 31, 2025, and the transcript captures the point clearly. (Apple Podcasts)
My take: we are heading in that direction. The slope varies by role, but the trajectory is obvious for most “screen jobs,” including big chunks of programming and digital ops.
What Musk actually said (context, not clickbait)
Digital vs. physical divide: AI “will take over” jobs that are purely digital; hands-on trades have more time. (Podcasts - Your Podcast Transcripts)
Timing and scale are debated, but direction isn’t. Musk has made similar claims before (e.g., “universal high income” era as AI replaces jobs). (New York Post)
This appearance is fresh: JRE #2404 (Oct 31, 2025) is the episode in question. (Apple Podcasts)
And Musk’s view isn’t isolated: Mark Zuckerberg said earlier this year that AI will eventually do all coding for Meta’s apps. Different voices, same vector. (Business Insider)
Is “all computer jobs” hyperbole? What current data shows
Exposure is massive: Goldman Sachs (2023) estimated up to 300M full-time jobs exposed to automation globally. That doesn’t mean instantly eliminated — but exposed. (goldmansachs.com)
Developers already use AI — but don’t fully trust it: The 2025 Stack Overflow survey shows 84% of developers use or plan to use AI tools, while trust in accuracy declined vs. 2024. Adoption up, skepticism up. (IT Pro)
Measured productivity gains exist: Controlled studies find ~20%+ time savings with AI assistance on coding tasks (effect sizes vary by task & experience). (arXiv)
Organizations are behind their workers: McKinsey (Jan 2025) finds most companies are investing in AI, but leadership/operating-model maturity lags; workers are often ready before process is. (McKinsey & Company)
Translation: The tools are here, velocity is real, institutions are catching up — and the more a job is purely digital and routine, the more it shifts from “do the work” to “supervise the work.”
For web developers: what actually changes
1) From typing code → orchestrating systems
Code-gen is table stakes. The advantage moves to spec clarity, system design, constraints, and tool/agent orchestration (retrieval, functions, data access, evals, deployment).
2) From features → outcomes
Clients don’t want “a React app”; they want conversion lift, speed, accessibility, analytics clarity, and AI-powered flows. Framing work in outcomes raises your value even if the code is machine-authored.
3) From individual productivity → pipeline productivity
Winning teams will standardize prompts, patterns, testing, guardrails, and CI for AI-generated changes, so shipping stays fast and safe.
4) From “code ownership” → data & domain ownership
If code becomes cheaper, proprietary data, integrations, and domain nuance become the durable moat.
Which programming tasks are most automatable (near-term)?
High-repeat boilerplate: CRUD scaffolds, API wrappers, tests from specs, CSS variants, component conversions.
Pattern-conforming refactors: Lints, migrations, typed infills, repetitive bug-classes.
Docs & glue: Typed signatures, comments, READMEs, SDK samples.
Stickier (for longer):
Ambiguous problem framing with incomplete requirements.
Security, privacy, compliance decisions and threat modeling.
Taste & product sense: UX trade-offs, brand feel, narrative.
Complex multi-system integration under strict SLAs.
The tasks shift first; the roles reorganize after.
A pragmatic playbook to stay valuable (and billable)
Own the spec. Learn to turn fuzzy business goals into tight PRDs, testable acceptance criteria, and prompts that produce minimal-surprise code.
Master agentic workflows. Don’t just “use a chatbot”; wire tool calls, evals, and rollback strategies so AI changes are testable and revertible.
Ship with guardrails. Policy, secrets, PII handling, rate limits, and automated security scans should be your default.
Measure outcomes, not hours. Commit to performance budgets, Core Web Vitals, A/B conversion metrics, and analytics hygiene.
Curate a personal model/tool stack. Know when to use fast vs. thinking models; keep a library of verified prompts, code action recipes, and evals.
Go up-stack: discovery workshops, information architecture, brand systems, copy frameworks — work AI augments but can’t cheaply replace.
Teach the machine. Build small internal datasets (component inventories, style tokens, API schemas) that make your AI outputs uniquely “yours.”
My bottom line
Musk’s claim is directionally right: computer-based work will be automated faster than physical trades, and programming is at the bullseye. But in the medium term, most dev jobs recompose before they disappear — moving from “typing code” to spec-setting, orchestration, and assurance. The winners aren’t the fastest typists; they’re the best problem framers, system designers, and outcome owners. (Podcasts - Your Podcast Transcripts)