An Incomplete Vision of Scince

Response to: https://scienceplusplus.org/metascience/

The intellectually generous catalog of reform ideas mistakes taxonomy for mechanism. Nielsen and Qiu are essentially saying: "imagine if there were many better ideas, and a healthy ecosystem tested and scaled them." That's not a theory of change; it's a description of what change would look like if change were already happening. The causal engine is missing.

Kanjun Qiu co-founded Astera, which funds exactly the kind of long-horizon, undirected science the essay advocates for—but through a completely opaque, invitation-only, founder-aesthetic-match process that makes the NIH look like open democracy. The essay explicitly criticizes "garage band research organizations that don't grow to worldwide pre-eminence" because there's "no natural feedback loop driving growth"—yet how does Astera solves this? A billionaire's taste and network. That's the kingmaker model with better aesthetics and a Substack voice.

Yudkowsky's interlocking equilibria framing maps cleanly here. The four bottlenecks Nielsen and Qiu identify (centralized control, collective action problems, network homogenization, no growth feedback for insurgent institutions) are real. But recognizing bottlenecks isn't the same as having a lever.

The essay gestures at "decisive metascientific results" as the lever—if we just had rigorous RCT-level evidence that Fund-by-Variance or Failure Audits work, institutions would adopt them. This is naive in exactly the way reform technocrats are always naive: NIH and NSF don't resist evidence-based reform because they lack evidence. They resist it because adoption would redistribute power, create accountability, and eliminate the reputational safety of consensus. Those are political facts, not epistemological ones.

The "metascience accelerator" they propose is implausible unless you control a large pool of unrestricted capital and can credibly threaten incumbent funders with irrelevance. The essay provides no account of where that leverage comes from.

The essay does usefully identify that the replication crisis shows that a sufficiently acute public legitimacy failure can temporarily suspend network effects and allow process change. That's actually the most actionable finding, and they don't press it hard enough.

Crystallizing ta public indictment requires exactly the kind of sustained, uncomfortable metascience entrepreneurship Nosek did—which means accepting that NSF, NIH, and VA will remain hostile until the crisis forces their hand.

The essay provids an interesting conceptual vocabulary, alas it is not a roadmap.