"After Automation" — @danshipper

Why this is in the vault

Shipper inverts the dominant "AI replaces jobs" narrative with a concrete operating thesis: cheap competence creates sameness ("slop"); humans stay structurally ahead as framers who decide which problems matter. Body did not render via Gmail (Every teaser only); reconstructed from canonical URL. This is load-bearing for the L4→L5 unhobbling trajectory and the cattle-shape specialist-fleet brainstorm — flag for manual review on first pass.

The core argument

• Cheap competence → sameness → demand for difference. Models train on visible expertise, making rare skills universally available. The system floods with default outputs. Humans reject the commoditization and demand outputs that feel "alive and specific, not cheap and generic" (verbatim, ≤15w).
• Frame vs framer. Models climb frames (predefined problem boundaries). Humans are framers — they decide which problems matter and reshape the frame as reality changes. When a benchmark saturates, the human moves the goalposts: scope-setting, risk assessment, migration strategy become the new task. The model must then climb the new frame. Shipper calls this "Zeno's paradox of AI."
• Smuggled intelligence. Hidden human judgment, feedback, and prompting embedded in benchmark setup makes model outputs look autonomous when they are not.
• Chart psychosis. Mistaking benchmark-curve extrapolation for actual future capability.
• Crucial caveat Shipper makes explicit: the pattern holds for expert-level knowledge work. He does NOT claim it explains the full economic picture for all job categories.

Every's concrete operating patterns

• Embedded agents, team-owned not personal. Shipper explicitly moves AWAY from "each employee gets one agent" because personal agents go stale without maintenance. Every runs a dedicated AI-engineering team that maintains the agent fleet.
• Named specialists: Fin (CS — closes 40.1% of conversations without human), Claudie (sales consulting — proposals/decks), Andy (editorial — collects "nuggets" from Slack), Viktor (general ops — metrics, surveys, memos).
• The human sandwich: Human frames task → AI executes → Human judges and extends.
• Senior Engineer benchmark (internal): model rewrites a vibe-coded production app from first principles. Human seniors score 80-90; GPT-5.5 scored 62/100. Prompt engineering (framing) dominates performance.
• Cost transparency: a single PowerPoint automation chain at Every includes "24 skills and 18 scripts" costing $62 per execution. Real operating cost of a specialist agent stack.

Mapping against Ray Data Co

The argument lands STRONG against four active threads. The frame-vs-framer distinction is the cleanest articulation yet of what Ray-the-COO is for:

• [[concepts/2026-05-20-rdco-specialist-fleet-brainstorm]] — direct corroboration. Shipper's "team-owned not personal" agent shift IS the cattle-shape architecture. Every is running the pattern in production with named specialists (Fin, Claudie, Andy, Viktor) and a dedicated maintenance team. Validates the founder's instinct that personal-pet agents stale out. Adds operating-cost ground truth ($62 per heavy execution) for services-pricing math.
• [[2026-05-18-indy-dev-dan-pi-to-pi-two-way-agent-orchestration]] — extends. Pi/IndyDevDan A2A patterns are the how (mechanical orchestration); Shipper's frame-vs-framer is the why (why a human still needs to be in the loop even when agents talk to each other). The framer role is what survives at L5.
• [[concepts/2026-05-20-services-pricing-model-for-rdco-future]] — calibrates pricing. The $62-per-execution datapoint anchors what a heavy RDCO specialist run actually costs to operate. Services pricing has to absorb that, plus the human-framer time.
• [[2026-05-21-stratechery-parag-agarwal-parallel-content-agentic-web]] — same week, complementary frame. Parag is on content valuation in the agentic web (supply side); Shipper is on human work in the agentic web (demand side). Both arrive at: cheap competence floods, scarce judgment differentiates.

Load-bearing question, answered: Does Shipper's argument support or complicate the Ray-the-COO thesis? Supports, with refinement. The thesis is NOT "Ray replaces the founder." It is "Ray is the cattle-shape specialist fleet; the founder is the framer." Shipper's frame-vs-framer language is sharper than RDCO's current vocabulary for the same idea. Candidate borrow.

Where this complicates the trajectory: the L4→L5 unhobbling is NOT a single agent getting smarter — it's the fleet getting smarter while a human framer stays ahead. That reshapes what "unhobbling" means. It is fleet-instrumentation + framer-tooling, not solo-agent-capability. Worth a follow-up concept note: "framer tooling vs operator tooling for RDCO."

Related

• [[concepts/2026-05-20-rdco-specialist-fleet-brainstorm]]
• [[2026-05-18-indy-dev-dan-pi-to-pi-two-way-agent-orchestration]]
• [[concepts/2026-05-20-services-pricing-model-for-rdco-future]]
• [[2026-05-21-stratechery-parag-agarwal-parallel-content-agentic-web]]
• [[2026-05-18-agentway-harness-engineering-claude-code-design-guide]]
• [[2026-04-20-indydevdan-pi-agent-teams-harness-engineering]]

Copyright note

Body did not render via Gmail (teaser-only newsletter format). Argument summary reconstructed from canonical URL https://every.to/p/after-automation via WebFetch. Verbatim quotes ≤15 words; remainder is paraphrase. Flag for manual review on first pass.