Approach

Our rules exist to stop us fooling ourselves

The same rules as our Indus script project, plus one more: every citation is checked, because AI agents invent references.

01The path

Every test is sealed before anyone sees the data

A claim counts only after five steps, in this order

  1. SealHash the plan and pass mark first. 5 sealed so far.
  2. PlantThe method must find a planted answer. The rule caught 28 of 28.
  3. NullCompare with random data that keeps the real data's shape.
  4. Red teamSomeone who did not build it tries to break it.
  5. PublishWrite it up either way, negatives included.
The path every claim takes, in order. A result counts only after the last step.

02Rules

Five rules we hold every test to

  1. Seal it before you look. Predictions, tests and pass marks are SHA-256 hashed before any result is seen. So is the rule for what makes two steps compatible.
  2. Use nulls that respect the data. Random networks keep each molecule's number of reactions; size-matched controls for assembly index; shuffled gene patterns for LUCA.
  3. Plant an answer first. A method must find a planted autocatalytic core, a planted ancestor or a planted incompatible step before its real output counts.
  4. Let someone else try to break it. Every positive result goes to a red team that did not produce it.
  5. Publish it either way. Failed routes are written up with the size of effect the test could have caught.

03Compatibility

A sealed rule decides when one step can feed the next

"Prebiotically plausible" is a judgement, so we fixed the judgement in advance. Step k+1 may use what step k leaves only through named physical processes from a closed list of 14, such as drying, heating, mixing or thermophoresis.

The rule then checks 9 hard fails and 3 warnings. Its tolerances, such as pH within 1 unit and temperature within 10 °C, are our own generous conventions, not literature values.

The rule fails an acid product poured into an alkaline step, and passes the same step once a real process fixes the pH

  1. Planted fault

    Step 1 leaves pH 2Same potStep 2 needs pH 10Flagged: F1 pH
  2. Twin

    Step 1 leaves pH 2Mix half and half with a pH 12 streamStep 2 needs pH 10Passes
Real data. The planted control the rule had to pass before scoring any real route. Source: experiments/compat/results.json

Result: 28 of 28 planted cases correct. Rule sha256 bfa572ed574fe6da..., file docs/prereg/compatibility_rule.md.

04Done

What "cracked" means, and how far we are

Two questions, kept apart: can chemistry do it at all, and what actually happened on Earth? "Cracked" means one stated setting and one route from its raw materials to a system that copies itself with heritable variation, where:

  1. Every step has a published yield from a lab other than us.
  2. Every step follows from the last by a named physical process, under a compatibility rule we hash in advance, with no purified hand-offs.
  3. The route predicts new data: a kinetic model fitted on the earlier steps predicts at least 3 measurements published after the hash.
  4. It survives a red team that did not build it.

One target of three met so far

Routes on the scored map4 of 4
Published claims benched against chance0 of 10
Sealed predictions confirmed by later data0 of 3
Real data from the live board. Bars run from zero to the target for the open goal, due by the end of 2027.

Our honest odds

under 1%Fully cracked by end of 2027
about 1 in 8The open goal: scored map, 10 claims benched, 3 predictions confirmed

What would raise them most:

  • New Bennu or Ryugu compound tables released after we seal predictions about them.
  • A lab reporting sustained exponential replication by a small ribozyme, which we model before the result is out.
  • One wet lab agreeing to run a test we pre-register.

05Traps

Five traps we watch for

  • Possible is not historical. Several routes may all work; that does not say which one happened.
  • 'Prebiotically plausible' is a judgement. So we fix the rule in advance.
  • Yields hide hand-offs. Most published steps start from purified chemicals; we mark every one.
  • Topology is not kinetics. A network that contains an autocatalytic set may never run.
  • Databases record modern life. Metabolic networks are biased toward today's enzymes; our nulls account for it.

06Not claimed

What we do not claim

  • That a working route is what happened. Showing a route is possible is not showing it is the one Earth took.
  • Wet-lab results of our own. We compute and reanalyse; only labs and sample-return missions can confirm.
  • Any citation we have not checked. Every reference passes a DOI audit before it appears here.

07Sealed

5 plans sealed so far

Each line of docs/prereg/HASHES.txt holds the SHA-256 of a plan, written before its data were looked at.

  • compatibility ruledocs/prereg/compatibility_rule.mdsha256 bfa572ed574fe6da...
  • assembly index size nulldocs/prereg/assembly_index_size_null.mdsha256 b60f5f12adf90262...
  • raf null xavier2020docs/prereg/raf_null_xavier2020.mdsha256 22e3ab3f6ec717ea...
  • qt45 net replicationdocs/prereg/qt45_net_replication.mdsha256 063f366d049ea216...
  • luca planted recoverydocs/prereg/luca_planted_recovery.mdsha256 4b4aff1067cc727e...