The Waymo Effect: How Frictionless AI Is Reducing Research Collaboration

The Waymo effect – a concise definition

The Waymo effect is the loss of valuable human friction when a technology makes interaction with other people optional, and researchers treat that convenience as a net gain.

The term comes from an essay by Daniel Hook, who noticed that a driver‑less Waymo ride feels like a personal oasis because it eliminates the need for small talk with a human driver. The same logic now applies to large language models (LLMs) that replace human collaborators.

Frictionless colleagues versus real collaborators

LLMs are always available, have no agenda, and only respond to the prompts you give them; human collaborators bring independent perspectives, challenge assumptions, and introduce serendipitous ideas.

A human co‑author may disagree, suggest a different framing, or point out a dead‑end you never considered.
An LLM will critique exactly what you ask it to critique, but it will not surface unexpected problems, prior failed attempts, or rival work unless you explicitly request that information.
The inconvenience of a human partner – negotiating author order, aligning schedules, managing ego – is actually the source of the collaborative value.

Decollaboration: the systemic incentive trap

Research incentives now reward speed and output, making it rational to replace costly collaboration with frictionless AI.

Three forces amplify this trend:

Funding cuts eliminate travel, workshops, and sabbaticals – the physical infrastructure of serendipity.
Velocity worship values rapid paper production; LLMs promise literature reviews and drafts in days instead of weeks.
Zero‑ego AI offers unlimited assistance without demanding credit, turning collaboration into a cost‑free shortcut.

Team‑science studies show that small, diverse teams generate disruptive breakthroughs, while large, homogeneous groups tend to develop existing ideas. Early evidence on generative AI mirrors this pattern: individual productivity rises, but the diversity of ideas narrows.

Writing is thinking – why removing friction hurts understanding

The act of writing forces thinking; outsourcing writing to an LLM removes the desirable difficulty that creates durable understanding.

Robert Bjork’s desirable difficulties research shows that effortful retrieval, delayed feedback, and struggle improve learning. When an LLM drafts a paper for you, you skip the cognitive work that reveals hidden gaps in your argument. As models become more fluent, the illusion of speed grows while the hidden loss of insight deepens.

Community insights from Hacker News comments

Consensus points

Convenient AI can erode interdisciplinary dialogue. Commenters note that researchers increasingly treat LLM outputs as “truth,” bypassing the critical exchange that would occur with domain experts. (e.g., @Frost1x, @abought)
Speed comes at the cost of understanding. Several users observe that AI‑generated code or analysis feels fast but often masks a lack of genuine comprehension. (e.g., @mccoyb, @xivzgrev)
The problem is not the tool but the incentive structure. Multiple comments echo the essay’s claim that institutions must fund and reward the friction of collaboration rather than only output metrics. (e.g., @chairleader, @gradus_ad)

Counter‑perspectives

Some argue that AI can enable new human interactions by providing a knowledge base that sparks conversations that would not otherwise happen. (@navaed01)
Others point out that forced small‑talk with drivers is not the sole source of outside perspectives, suggesting the “last stranger” claim may be overstated. (@crazygringo)
A few note that similar dynamics have appeared in open‑source software: code contributions explode while community engagement stalls. (@maddiaa0)

Pilot‑in‑command science – a proposed remedy

Researchers should remain the “pilot” while AI agents act as “crew” that assist but do not steer the research direction.

Dashun Wang (Nature, March 2026) advocates pilot‑in‑command science: the human defines the question, the AI drafts, critiques, and plans, but the researcher retains authority over conclusions. To preserve diversity, Wang recommends cultivating multiple models with differing reasoning styles, effectively re‑introducing engineered friction.

Funding the friction that sustains collaboration

Institutions should treat collaborative activities—workshops, visits, unstructured discussions—as essential infrastructure and evaluate researchers on who they think with, not just on publication velocity.

Allocate dedicated budgets for travel, sabbaticals, and co‑location.
Incorporate “collaboration impact” metrics into hiring, promotion, and grant reviews.
Require statements such as “who did you think with?” alongside traditional impact summaries.

Final takeaway

The Waymo effect warns that the convenience of AI can silently turn researchers into passengers in a smooth, solitary ride; the crucial question for research leaders is who is in the front seat and how we can fund the friction that keeps that seat occupied.