Life uses a tiny fraction of the chemistry available to it. This chain is entirely open at every step — and it's the best entry point here for a technical reader, because it contains a live experiment anyone could run.
Entirely open.
The number is startling, and it isn't this framework's — it's already in the literature, unframed.
Wołos et al. (Science, 2020) supply it: from six substrates under 614 transforms to generation 7, the network reaches 82 biotic molecules and 36,603 abiotic ones. Life uses roughly 0.2% of what's reachable.
→ Is the realised set distinguished by being cheapest, or by something else?
Redundant — on their own data.
Removing 34 of 63 reaction classes still leaves every biotic molecule synthesisable via bypass routes — against only 8 removable classes for the abiotic set. Biotic molecules are roughly four times more robust to rule deletion.
Wołos et al. attribute the asymmetry to scale-free architecture and preferential attachment — but scale-freeness is a property of the whole network, and biotic and abiotic molecules share the same graph, the same hubs, the same topology. Architecture can't explain why one subset is four times more robust than the other. They report the asymmetry and leave it unexplained.
→ Is the asymmetry real, or an artefact of what biotic molecules happen to be like?
Possibly — and this is genuinely unresolved.
The paragraph immediately above their own result says biotic molecules are more thermodynamically stable, more hydrophilic, carry fewer functional groups — 2.67 against 3.52 — and are reachable with fewer changes of condition. Simpler molecules have more routes almost by construction.
Is "all 82 biotic molecules remain synthesisable" even a fair comparison to "all 36,603 abiotic molecules remain synthesisable"? Keeping a small set accessible is easier than keeping a large one — the two robustness figures may not be measuring comparable things.
→ Does the biotic/abiotic robustness asymmetry survive matching on functional-group count, molecular mass, and network generation? Survives → the redundancy claim holds on independent data with the confound removed. Vanishes → the claim loses its best evidence, which is worth knowing and is exactly the kind of thing nobody publishes because nobody looks. Either result is publishable, and it runs on data that's already public — no new chemistry needed.
No paper of this framework's own yet. Everything above is a reading of Wołos et al. 2020 plus one unpublished observation about what their own architecture-based explanation can't account for.
Open, at every step — which is why this is the best entry point here for a reader who wants to check the framework's actual working, not just read its conclusions. See /open/.