Cheap Competence, New Frontier — Every Context Window (2026-05-24)

Hybrid issue: Dan Shipper lead essay framing the week's theme, plus curated items from Every staff (Jack Cheng on Gas City multi-agent factory, Alex Duffy on Google I/O, Katie Parrott on entry-level career pivots, Ashwin Sharma on physicians) and a footer pointer to "After Automation."

Core thesis (Shipper)

Automation does not contract the surface of human work — it expands it. As AI commoditizes a layer of competence, a new frame opens above for humans to define, oversee, and orchestrate. "There's always a new frame for humans to hand the models." Pattern repeats across:

• Every's own ops: agents do routine work; demand for strategic judgment rises, not falls
• Anthropic acquiring Stainless for $300M: an agent can't use a company's API unless a human first made it easy to use. Framing is the moat, execution is commodity
• Google I/O 2026: model benchmarks now table-stakes; the keynote story shifted to distribution (Gemini 3.5 Flash at "Opus 4.7-level intelligence, 4× speed, half cost")
• Gas City (Jack Cheng preview): a "mayor" agent dispatches anonymous worker agents — human orchestration is the scarce skill in the 100-agent factory pattern

Secondary thread: doctors as canonical example

Ashwin Sharma's column: half of US consumers now self-diagnose via LLMs, so medical knowledge is commodity. But patients arrive with sharper questions, and the bottleneck moves to situated judgment — knowing what to do next for this particular person. "AI does not make physicians irrelevant. It just makes excellent physicians more valuable." Mediocre middle compresses, top tail expands.

Other curated items

• Katie Parrott on starting a career when AI is doing the entry-level job — 22-to-25-year-olds in AI-vulnerable roles saw employment drop 13% since late 2022; includes a copy-paste career-coach prompt
• Mike Taylor on Google I/O takeaways ("agents, agents, agents")
• Executive AI Sessions webinar (June 2) on leadership adoption

Why this is in the vault

Three reasons:

1. Direct anchor for the L5 thesis. Founder is at L4 actively building toward L5, focused on unhobbling the COO agent (toolset + visibility) rather than operating bets first ([[project_l5_north_star_strategic_direction]]). Shipper's "new frame for humans to hand the models" is the cleanest articulation yet of why Ray's capability ceiling is the bottleneck, not the bet inventory.
2. Stainless framing as moat. The $300M Anthropic acquisition is a $-priced bet that "human framing of integration surface" is durable value. RDCO's targeting-system filter ([[feedback_targeting_system_prioritization_filter]]) is the founder's local version of the same insight — niche + bottleneck framing precedes any agent execution.
3. Doctors-as-template generalizes. The "commodity middle compresses, excellent tail expands" pattern is directly applicable to Sanity Check's positioning: the newsletter has to be original re-frame, not summary ([[feedback_no_derivative_sanity_check_pieces]]) — i.e. occupy the "excellent" tail, because the commodity middle is now AI-generated for free.

Mapping against Ray Data Co

• L5 capability build (RDCO COO agent): Shipper's frame says capability gain doesn't reduce founder work, it just shifts it up the stack. Validates the order-of-operations: unhobble Ray first, then bet inventory grows naturally. Counter-evidence to any "just operate small bets first" critique.
• Sanity Check positioning: "Excellent physician" mental model = Sanity Check has to be the re-frame layer above commodity AI summaries. Specifically rules out (again) any thread of "let Ray summarize what others wrote." Re-confirms [[feedback_no_derivative_sanity_check_pieces]].
• MAC info-product targeting: Stainless = "integration surface that makes the API usable" is exactly the MAC value prop framing — buyers don't need more data, they need someone who's already framed the integration. Worth re-checking MAC landing-page copy against this lens.
• Investing thesis hooks: "Cheap competence" + "Anthropic buying integration moats" = candidate research-backlog item — which incumbents own irreplaceable integration surface that an agent layer cannot disintermediate? Worth queuing to /curiosity.
• Hiring/contractor frame (forward-looking): The 13% entry-level employment drop is the wedge for any future "we hire only senior + agents" RDCO position; not yet relevant but worth tagging.

Verdict

Strong. Shipper essay is high-signal and directly load-bearing for L5 strategy. Doctors example is reusable across multiple RDCO surfaces. The Stainless framing is investing-thesis raw material. Hybrid issue — lead essay is the keeper, secondary items are useful pointers but the Shipper piece carries the value.

Related

• [[2026-05-21-every-after-automation]] — referenced in this issue's footer; the longer-form Shipper essay this newsletter distills
• [[2026-05-13-zoharatkins-jevons-paradox-torah-learning-cheap-knowledge-insight]] — Jevons-paradox framing of cheap knowledge expanding demand for higher-order interpretation; tight analog to the "doctors" thread
• [[2026-02-23-every-chatgpt-memory-context-rot]] — earlier Every piece on memory/context as the scarce primitive when models are abundant
• [[project_l5_north_star_strategic_direction]] — RDCO position the essay anchors
• [[feedback_no_derivative_sanity_check_pieces]] — operational rule the "excellent physician" thread reinforces
• [[feedback_targeting_system_prioritization_filter]] — local-to-RDCO restatement of the Stainless framing-moat insight