Back to Blog
vibe codingfuturevibecoderai future

Is Vibe Coding the Future? A Sober 2026 Take

Vibe coding shipped real businesses (Lovable's $200M ARR) but Andrew Ng and Simon Willison are right that the framing is overhyped. The honest 2026 read — where vibe coding is the future, where it is not, and what changes by 2028.

··7 min read

Is Vibe Coding the Future? A Sober 2026 Take

TL;DR

  • Yes for the build phase. No for the operational phase, at least not yet. The honest answer is "partially, and the partial-ness matters more than the headline."
  • Lovable's $200M ARR proves the build acceleration is real. Andrew Ng's and Simon Willison's critiques prove the framing overstates the engineering displacement.
  • What changes by 2028: agentic systems will close more of the operational gap. What stays the same: the human judgment for security, scale, compliance, and product decisions.

The "is vibe coding the future" thread shows up monthly on Hacker News, Indie Hackers, and X. The takes are usually binary — full evangelist or full skeptic. The honest 2026 answer is more interesting and less satisfying than either pole. This cluster sits inside our vibecoder distribution playbook and pairs with vibe coding hangover on the operational reality check.

The case for "yes, the future"

Three durable pieces of evidence:

1. The build acceleration is real and meaningful. Per TechCrunch's reporting on March 6, 2025, 25% of YC's Winter 2025 cohort had codebases that were 95% AI-generated. YC managing partner Jared Friedman: "It's not like we funded a bunch of non-technical founders…A year ago, they would have built their product from scratch — but now 95% of it is built by an AI." The build phase compressed from months to weeks to days for a meaningful share of new startups.

2. The economics of a vibe-coded company are different. Lovable, per TechCrunch (July 23 and November 19, 2025), hit $100M ARR in 8 months, $200M ARR four months later, raised $330M at $6.6B in December 2025. CEO Anton Osika told Slush 2025 the company reached $100M ARR "faster than OpenAI, Cursor, Wiz, and every other software company in history." These numbers are not possible under the previous build-cost regime.

3. The audience is large and growing. Collins Dictionary named "vibe coding" Word of the Year on November 6, 2025. Cursor, Claude Code, Lovable, Bolt, Replit, v0 all grew at compounding rates through 2024-2025. The category is not a fad — it is a sustained productivity shift that the broader software industry is internalizing.

The case for "no, overhyped"

Andrew Ng and Simon Willison have voiced public skepticism. The substance of the critique:

1. The "vibe" framing understates the judgment required. "Just vibe with it" is misleading shorthand. Production software requires architecture decisions, security review, scaling design, and ongoing operational work. Vibe coding tools do not yet supply any of these. The framing implies the engineering is solved; the reality is the engineering shifted from the build phase to the operational phase.

2. The failure modes are systematic. Vibe-coded apps consistently ship with security gaps, scaling fragility, and unmaintained code structures. The full reckoning is in vibe coding hangover. The pattern is so reliable that it is now a recognized stage of the vibecoder lifecycle, not a personal failure.

3. The aesthetic dominance creates false confidence. Lovable apps look polished from day one. The polish hides architectural fragility that takes weeks of usage to surface. Founders consistently mistake aesthetic completeness for operational completeness, then discover the gap during their first scaling event or first security incident.

Both critiques are honest and substantive. Dismissing them produces founders who hit the hangover at full speed.

The sober integration

The honest 2026 read integrates both:

  • The build phase: vibe coding is the future. Founders not using AI-native tooling are operating at 1/10th the velocity of founders who are. Within 2-3 years this gap will compound into existential disadvantage for non-AI-native teams.
  • The operational phase: vibe coding is not yet the future, but it is moving that direction. Agentic systems for security review, scaling, compliance, and operational hardening are emerging but not yet at production parity with human engineering judgment. Founders running vibe-coded apps in 2026 still need to do the operational engineering themselves.
  • The product phase: not affected by vibe coding either direction. What to build, who to build for, what to charge — these are still human judgment calls. AI helps with the analysis; the decisions remain human.

The summary: vibe coding is the future of building software, partially the future of operating software, and not really the future of product decisions. Three phases, three different answers, all true simultaneously.

What changes by 2028 (best-guess)

If the 2024-2026 trajectory continues:

  • Agentic systems will close more of the operational gap. Tools for AI-driven security review, performance optimization, and compliance scaffolding will mature. The operational engineering required of a vibe-coded app founder will decrease.
  • The build acceleration will plateau. The 10x speedup from manual coding to vibe coding will not be followed by another 10x. We will be on a slower productivity curve at the post-vibe-coding baseline.
  • The competitive landscape will saturate. As vibe coding tooling becomes universal, the differentiation moves to distribution, brand, and customer relationships — exactly where it has been for non-tech startups for decades.
  • The terminology will shift. "Vibe coding" was coined February 2, 2025 (Karpathy, 6:17 PM); by 2028 the term may feel dated. The underlying practice (AI-native software development) will persist.

What stays the same: human judgment for product decisions, customer relationships, brand voice, security trade-offs, compliance scope, and the broader question of what to build.

What this means for builders in 2026

Practical implications:

  • Use vibe coding tooling aggressively for the build phase. Founders not using Cursor / Claude Code / Lovable / Bolt are leaving 5-10x velocity on the table.
  • Plan for the operational phase honestly. Budget 2-3 months for production hardening after launch. Do the security audit. Plan the rewrites of load-bearing components.
  • Invest in distribution proportionally. Because building got cheaper, the marginal investment in distribution is now higher than the marginal investment in features. Most founders under-invest in distribution and over-invest in product.
  • Stay humble about the framing. When operators ask if vibe coding is the future, the honest answer integrates both the proof points and the critiques. Founders who only carry the optimistic framing get caught by the hangover; founders who only carry the skeptic framing get out-shipped.

Sibling clusters

FAQ

Is Andrew Ng's critique a reason to not use vibe coding tools? No. The critique is a reason to acknowledge the operational phase exists and plan for it, not to skip the build acceleration. The math still strongly favors using AI-native tooling for the build phase; the critique applies to what comes after.

Will vibe coding tools eventually do the operational work too? Trending that direction but not yet at production parity. Anthropic, OpenAI, Cursor, Anthropic Claude Code, and several startups are working on agentic systems for ops, security, and scaling. By 2028 the gap will be smaller; by 2030 it may be largely closed. In 2026 the gap is real and the founder still needs to absorb it.

Should I avoid the "vibecoder" label because it might age poorly? Up to you. The label is currently a marketing asset (small audience for the term, high engagement when claimed) and may be a liability later (terminology fades). For 2026-2027 the label is strongly net-positive for distribution. By 2028 reassess.

What changes for non-technical founders specifically? The build phase becomes accessible (Lovable, Bolt enable non-technical builds), but the operational phase becomes a harder constraint than for technical founders. Non-technical founders without engineering partners run into the hangover earlier and with fewer tools to address it. The honest framing: vibe coding lowered the build barrier but did not eliminate the need for engineering judgment somewhere in the team.

Is the YC W25 95%-AI-codebase figure exaggerated? Per TechCrunch's reporting on March 6, 2025, the figure comes from Jared Friedman's remarks in a YouTube video, not a formal YC publication. The 95% figure is per-startup, not aggregate. Cite as "per TechCrunch reporting Friedman's remarks" rather than as institutional YC data. The directional claim (large portion of YC W25 codebases AI-generated) is robust; the precise 95% figure deserves caveats.

Building is no longer the bottleneck. Visibility is. buildinpublic.so is narrative infrastructure that runs inside your building workflow — for the build acceleration that vibe coding produces, the marketing infrastructure has to compress at the same rate. Dev Cards, Loudy, and Vibey close the gap between shipping fast and distributing fast.