Why this is in the vault
Beach reframes "AI code is cheap" as a power-dynamics claim rather than a cost claim — worth keeping as the sharpest one-line rebuttal yet to the flat "tokens are cheap so it doesn't matter" argument RDCO hears from vendors and executives alike.
The core argument
The post is paywalled past the opening section, so what's filed here is the newsletter's rendered thesis paragraph, not the full essay — flagged via source_fidelity rather than filed silently as complete.
Beach's opener: "The vibes are cheap; even if the per-token cost increases 10x or 100x, it doesn't matter, the vibes are cheap; at least they are cheap from the viewpoint of those who hold the power. The power to hire and fire, to set technical direction, and to point a finger and send the peons running." The move is deliberate: he separates the actual per-token cost of AI-generated code (which can rise 10-100x and genuinely matters to whoever pays the bill) from its perceived cost to the people making the adoption decision. For an executive who doesn't carry the maintenance burden or the blame when something breaks, "cheap code" isn't a statement about compute economics at all — it's a statement about who absorbs the downside. If a vibe-coded system fails, the fix is to redirect blame downward ("send the peons running"), not to re-litigate the decision to generate code cheaply in the first place. That asymmetry — decision authority and cost exposure sitting with different people — is the polemic's target, not the technology itself.
Mapping against Ray Data Co
Beach's asymmetry ("the power to hire and fire... to point a finger") is the inverse of the founder's own posture on AI-coding cost, captured in feedback_api_cost_budget_controlled: Ray runs on a Max subscription with metering invisible to the founder, and the explicit rule is never report dollar figures because the founder — like the executives Beach is describing — isn't the one who needs to feel per-token cost pressure to make good decisions. The founder inverted the dynamic Beach criticizes: rather than letting "cheap from the viewpoint of power" produce blame-shifting when something breaks, RDCO's IC-mode-vs-production-mode split (feedback_ic_vs_production_mode) puts the real cost-of-being-wrong check on the output (does it ship to something public-facing) rather than on who's watching the meter. It's also a direct companion to two other Beach pieces already in the vault on the same throughline — token-cost economics in [[2026-06-11-data-engineering-central-agentic-token-tax-opencode-ollama]] and skill erosion under heavy agent use in [[2026-08-03-data-engineering-central-keeping-technical-skills-llm-age]] — all three converge on the same worry from different angles: technical (token tax), personal (skill atrophy), and now organizational (who owns the failure when the code is "cheap").
Related
- [[2026-08-03-data-engineering-central-keeping-technical-skills-llm-age]]
- [[2026-06-11-data-engineering-central-agentic-token-tax-opencode-ollama]]
- [[feedback_api_cost_budget_controlled]]
- [[feedback_ic_vs_production_mode]]