Head to head, honestly
Independent vs institutional alignment research
Not a superiority claim in either direction. The frontier laboratories employ the strongest alignment researchers in the world, publish work this programme learns from, and disclosed their own containment failures themselves. The comparison is about structural position: which experiments are cheap from where you stand.
What position shapes
Inside a laboratory, the instrument belongs to the party being measured: the evaluation runs inside the organisation whose product is being evaluated, which is not corruption but is a hard place to stand. A negative result about your own model is a difficult publication and an easy deprioritisation. A rebuild of the objective function is a product decision competing against a release schedule. And an architecture that must be adopted at genesis is close to impossible to test in a system that already has users.
What independence buys, and what it costs
The experiments that are structurally awkward inside a laboratory are cheap for someone with nothing to protect: the ones that risk your own headline number, the ones that require blinding your own scorer, the ones that only make sense before a system is built. This programme has run exactly those, including the attack on its own instrument that reversed a result's sign under blinding, and the retraction of its own most striking number. The costs are stated with equal plainness: no institutional peer review, single-laboratory results, no independent replication yet, and resources that do not compare. Independence is a vantage point, not a credential.
Complementarity, not competition
The honest model is division of labour. Institutional research has the scale, the systems and the strongest people; independent research can afford the self-damaging experiment and the genesis-time question. The replicate page publishes exact commands so that anyone, institutional or not, can rerun what this programme claims; a refutation gets published with the refuter's name on it. That standing invitation is the whole relationship this page proposes.
Sources: the open problem page; the corrections log; the replicate page; how to weigh work of this kind.
A longer essay on the same ground: The future of independent AI alignment research.