A16z partner says that the theory that we’ll vibe code everything is wrong

a16z, VCs, and Motives

• Many commenters dismiss the firm’s opinions as PR for existing AI/LLM bets, citing past crypto promotion and “pump-and-dump” behavior.
• Some argue VCs don’t need to be right, only to expose LPs to risk and generate deal flow; insight is secondary.
• Others note this is a single partner talking, not the entire firm, and criticize ad‑hominem dismissal without engaging the argument.

Fascism, Marinetti, and AI as “Fascist Technology”

• A tangent debates the firm’s choice of a historical “patron saint” linked to the Fascist Manifesto.
• One side stresses that early fascist texts included progressive‑sounding planks (universal suffrage, 8‑hour day, wealth tax) and that “fascist” as a label has drifted.
• Others counter that the author remained an ardent fascist supporter in practice, so the symbolism is still damning.
• Another thread ties LLMs’ plausibility, opacity, and centralizing tendencies to arguments that AI is structurally hostile to democratic institutions.

Vibe Coding: Hype vs Reality

• Strong skepticism that “we’ll vibe code everything” (natural-language prompts replacing software engineering) will happen soon, if ever.
• Dijkstra’s critique of “natural language programming” is invoked: ambiguity is intrinsic to human language, while software needs formal, unambiguous representations.
• Several argue AI won’t replace domain experts/SWEs but may suppress junior hiring and hollow out the talent pipeline.
• Others think “vibe coding” is already the “new normal” for everyday work, even if it makes programming less enjoyable.

Where AI Coding Helps (and Fails)

• Supporters highlight AI’s strengths: rapid iteration, translation tasks (e.g., x86→ARM), boilerplate, wireframes, alpha prototypes, and working against strong test harnesses.
• Critics describe fragile “vibecoded” deployment scripts and systems that appear to work but are undocumented, non‑deterministic, and hard to maintain.
• Some claim modern agents already review and test code better than humans; others respond that correctness relative to fuzzy business requirements is still a human bottleneck.

Build vs Buy, SaaS, and Enterprise Software

• One camp expects AI to make in‑house clones of tools like Jira, Linear, CRMs, or e‑commerce platforms increasingly attractive, especially given SaaS bloat and high prices.
• Opponents stress hidden costs: ops, security, compliance, backups, and long‑term maintenance; they predict SaaS will cut prices rather than be wholesale replaced.
• Examples are given of selectively rebuilding generic tools (task tracking) while avoiding anything touching legal/financial infrastructure (payroll, banks, cap tables, signatures).

Broader Role of AI

• Several argue AI’s real power is as an exploratory/learning tool and knowledge engine, not a one‑shot code generator.
• Others speculate on macro effects: AI eroding labor advantages, changing capital moats, and enabling “single‑player” engineers to rival large teams.