Peter Kahl’s article reframes distributed cognition not as an automatic virtue of decentralised intelligence, but as a fragile epistemic infrastructure whose reliability depends on governance, incentives and closure mechanisms. The central critique is that contemporary discourse too often treats prediction markets, open-source communities, digital platforms, deliberative bodies and regulatory institutions as interchangeable expressions of “crowd wisdom”, despite their radically different architectures. Kahl therefore distinguishes between systems that distribute probabilistic belief, cognitive labour, judgement, attention or delegated authority, showing that each stabilises knowledge through a different locus of epistemic closure. Prediction markets, for example, convert dispersed beliefs into price signals, but remain vulnerable to capital-weighted narrative capture; open-source communities distribute problem-solving labour through shared artefacts, yet may consolidate authority in maintainer oligarchies; platforms rank visibility algorithmically, often amplifying salience rather than truth. The article’s case-study contrast between prediction markets and open-source software demonstrates that decentralisation does not abolish power: it merely relocates it into prices, code, procedures, rankings or institutional decisions. Its decisive contribution lies in replacing naïve enthusiasm for collective intelligence with an architectural account of epistemic reliability. Distributed cognition, on this reading, succeeds only when contestability, transparency, incentive alignment and error correction are deliberately designed. The conclusion is therefore austere but generative: collective knowledge systems should not be celebrated because they are distributed, but evaluated according to how responsibly they govern what becomes credible, salient and settled.