How did sophisticated, supposedly factor-neutral teams at @JaneStreetGroup, @WeAreHRT, and many other highly capable firms lose money together? We present an explanation (after dozens of conversations with major market makers, stat-arb shops, semi-systematic desks, and discretionary long-short pods) connecting 3 ideas: 1) time-varying factor exposures 2) signal alignment induced crowding 3) increasingly similar research processes This is a working hypothesis and we may well be wrong here. First, we should distinguish a factor premium from a portfolio’s current factor exposure. A factor premium is the expected compensation associated with systematically holding a particular kind of exposure over time. Your current factor exposure is how correlated the positions you hold today are to the set of holdings currently associated with a factor. A strategy can have little full-sample correlation with momentum or value while carrying substantial exposure to either at particular moments (e.g. a factor-timing strategy). Low historical correlation is not the same as continuous neutrality. Sophisticated firms generally know this and often constrain current exposures directly. The deeper issue is that neutrality is always relative to a model i.e. a portfolio can be neutral to the market, sectors, and standard style factors while remaining exposed to an unmodeled theme. Accurately hedged conventional factors do not eliminate risks to unmodeled factors / crowding risk. Now consider moments of signal alignment. Suppose earnings revisions, price momentum, alternative data, and fundamental research all favor the same AI-related stocks. Agreement can rationally increase confidence, provided those signals contain sufficiently distinct information. Yet that agreement also tells you something about everyone else. If competing firms use imperfectly overlapping signals, a stock that looks attractive through many different lenses may now appeal to a much wider set of investors. In addition, when signals align, firms running strategies with time-varying factor exposures may be more likely to converge on the same exposures at the same time as well. Your strongest consensus trade therefore may also be the trade most likely to appear in everyone else’s portfolio as different research processes can converge on the same holdings. Signal agreement can therefore increase both confidence in the forecast and exposure to crowding. The first effect supports the trade; the second can reduce its remaining return and increase its vulnerability to an unwind. This is also a difficult interaction to estimate as each signal may have substantial historical evidence behind it, but separately validating several signals does not establish the payoff from their joint occurrence, especially after accounting for how other investors respond to that same combination. The problem is not that sophisticated firms never model interactions but that rare combinations offer fewer independent observations. Also, an alignment that worked when little capital followed it may behave differently after widespread adoption (which is made possible by AI). When an unwind begins, those shared positions can transmit losses across apparently different strategies. Funds sell overlapping longs, cover overlapping shorts, and reduce risk together, and positions that looked diversified under ordinary conditions become connected through liquidity constraints. AI-assisted research intensify this convergence. If many teams leverage similar foundational models trained on overlapping information and exhibiting similar analytical tendencies, some apparent diversity of research may conceal shared assumptions. In a limited sense, firms may increasingly be “hiring the same analyst / researcher.” I would however NOT conclude that this makes sophisticated trading obsolete. Firms that can distinguish independent information from shared consensus, estimate crowding, and develop differentiated forecasts will ALWAYS have an advantage. Teams like Jane Street and HRT have considerable talent to bring to that problem. That gives me confidence in their capacity to adapt, without assuming that intelligence alone makes any firm immune to a crowded unwind.