06-reference

every 100 thinkers human work after automation

2026-08-18·reference·source: Every·by Dan Shipper
future-of-workai-automationagent-economyharness-engineering

What Does Human Work Look Like After Automation?

Why this is in the vault

A 100-thinker roundup on what human work looks like once AI absorbs execution — directly upstream of RDCO's "agent-deployer" positioning and the L5 north star framing.

The core argument

Dan Shipper (Every) frames the question of the moment: as intelligence becomes abundant and execution gets automated, where will humans find meaning, status, and purpose at work, and what skills will still matter? His premise is that mainstream discourse doesn't have the answer yet, but the people living daily with frontier models — builders and operators, not pundits — do. He recaps Every's own running thesis since the GPT-3 era: one person can now do team-scale work, knowledge workers become "managers of models," and automation paradoxically creates more work for human experts rather than less.

The piece then opens with six of the 100 contributors' one-line "shots" on the future of work:

Mapping against Ray Data Co

The "knowledge workers become managers of models" framing is the same claim underlying RDCO's L4→L5 bet (project_l5_north_star_strategic_direction) — bets are downstream of agent capability, and the founder's own COO setup (Ray) is a live instance of "one person doing team-scale work" via delegated sub-agents. The Le Cunff line — "questions will become the hard part" — sharpens the targeting-system filter already in use (feedback_targeting_system_prioritization_filter): the constraint shifts from execution capacity to knowing which question to aim the agent fleet at. Worth flagging as a tension too: Every's own thesis that automation creates more expert-level work cuts against a pure headcount-replacement reading of agent economics — useful counterweight when framing phData or RDCO capability pitches.

Curation section

The article's spine is 100 short "shot" statements from named operators/thinkers (six sampled above from the accessible portion); the remainder is a long-form list format rather than external links, so no third-party curation links to evaluate under the deep-fetch cap. Zero deep-fetches triggered — the piece is self-contained commentary, not a links digest.

⚠️ Sponsorship

Gmail delivery rendered only a metered-paywall stub (subject teaser + "Start free trial" CTA + Every's own subscription-bundle upsell for Sparkle/Cora/Spiral/Monologue — house self-promo, not third-party). The fuller read via WebFetch of the canonical URL surfaced a sponsor mention for Svix (webhook infrastructure) inside the full article; this reconstruction could not independently confirm placement type (pre-roll block vs. in-narrative) since the paywalled Gmail body never rendered it directly. Treat as disclosed-but-unverified pending a future issue that confirms the pattern.

Related