Estimated energy usage:

What's going on?

There are two template-replicating polymers present from the start. Hide the bath to see them better. The two are: eacbdf (purple-red-green-blue-yellow-cyan) and ebdbdf (purple-green-yellow-green-yellow-cyan). They replicate by pulling in matching atoms from the surrounding bath.

The colors (abcdef) of the atoms never change. The numbers/letters you see (01233456789ABCDEFHI) are their states. We use the states to direct the replication process. Replication works like this:

eabcdf --> eabcdffcdbae --> eabcdf + fcdbae

Reactions happen between pairs of atoms, and can bond or unbond them and change their states. All reactions are reversible with equal probability. Bonding reactions are energetically downhill, unbonding reactions are uphill.

The reactions in our system are listed below. They are designed to let any sequence e{abcd}*f replicate. Credit to Dave Mann for inventing this form of template-replication (see the Organic Builder paper).

    
    x1yD    -> xDy1    // message-passing
    x2yD    -> xDy2
    x3yD    -> xDy3
    x4yD    -> xDy4
    x5yD    -> xDy5
    x6yD    -> xDy6
    x7 + a0 -> xDaE    // bond the appropriate type
    x8 + b0 -> xDbE
    x9 + c0 -> xDcE
    xA + d0 -> xDdE
    xB + e0 -> xDeF    // (if bonded an 'e' then we start reset1 (F) )
    xC + f0 -> xDfE
    x1yE    -> xDy7    // activate for the next one to be bonded
    x2yE    -> xDy8
    x3yE    -> xDy9
    x4yE    -> xDyA
    x5yE    -> xDyB
    aFxD    -> a1xF    // reset1 passes over a base 'a', setting its state to 1
    bFxD    -> b2xF
    cFxD    -> c3xF
    dFxD    -> d4xF
    eFxD    -> e5xF
    fFfD    -> fC + fG  // unbonds when reset1 passes over the central f-f
    fGxD    -> fGxH    // reset2 (G) starts moving
    aHxD    -> aGxH    // reset2 passes over a base
    bHxD    -> bGxH
    cHxD    -> cGxH
    dHxD    -> dGxH
    eHxG    -> e5xI    // reset2 ends, starts reset3 (I)
    aIxG    -> a1xI    // reset3 passes over a base
    bIxG    -> b2xI
    cIxG    -> c3xI
    dIxG    -> d4xI
    fIa1    -> fCa1   // reset3 gets to the end
    fIb2    -> fCb2
    fIc3    -> fCc3
    fId4    -> fCd4
    fIe5    -> fCe5

This is a bit complicated, which is annoying. Here are some notes:

states 1-6: passive message saying we would like to bond a-f respectively
states 7-C: active end of the growing molecule, ready to bond the next atom (of type a-f respectively)
state D: empty space in message-passing
state E: active end ready to be told what the next atom to be bonded is
states FGHI: resets 1,2,3,4

The simplest possible replicator is e5fC.