Parallelism doesn't beat Amdahl's Law. This is the reason simulations requiring a supercomputer never make their way into laptop CAD applications.
Bootstrapping is a sequence of phases, where you use the speed & parallelism of the current phase to fabricate systems that eliminate the current biggest bottleneck.
We need to perform sequential chemical steps:
- Fast (>1 Hz frequency)
- With high yield (>99%)
Yield limits the building block count of systems you can fabricate. A quick estimate is "complexity = 1 / (1 - yield)".
Limiting sequential frequency bottlenecks of atom placement:
Phase I:
- 18–8640 reactions/day, massive margin of uncertainty
- Bottleneck is detecting new pre-charged tips in inverted mode
- Conventional mode mechanosynthesis is a clever attempt to ignore the problem
- Incremental path is a clever attempt to jump to phase III
- Target frequency is ~1 Hz reaction rate, as good as the mammalian ribosome
Phase IV:
- Throughput is (gas molecules per perfectly sealed container) / (UHV bakeout latency)
- Tradeoff: larger containers store more molecules, but require more latency in Phase I to fabricate
- Requires seamless covalent welding to create perfect vacuum seals
- Requires molecule sorting rotors with quick binding/diffusion, 100% purity, no leakage
- To fabricate this, we need "part assembly" techniques from Phase V. To connect gas canister fragments fabricated in parallel on different Phase I workstations.
Phase III:
- Eliminate the latency to take 1 single reading from a macroscale analog sensor (1 Hz – 1 kHz)
- Make reactions that succeed 99.9999% of the time, removing the need to sense which product was built
- Enable the use of minimally compact replicators (like KSRM) in a distant future phase
Phase X:
- Eliminate the bottleneck of 1 UHV bakeout per gas canister refill
- Shrink the entire UHV SPM system from ~1 meter scale to ~1 μm scale, using atomically precise vacuum system parts
- Continuous flow of liquid feedstock from external canister
Somewhere in the timeline:
- Use nanoscale 3DOF actuators made of molecular machine parts
- Higher resonance frequency (MHz) than macroscale piezos (kHz)
- Smaller volume = more workstations can be packed into the same volume
Nanosystems Chapter 15 lays out the speeds and yields of biological machinery. I augmented these results with some from the Internet. Double check all the numbers and don't quote me on them.
DNA replicase might have inflated yields, because it can utilize error correction.
In practice, people use non-biological means like solid-phase peptide synthesis to make artificial polypeptides. It's more convenient to extract & purify than peptide chains stuck inside of bacteria. Plus, you can easily incorporate non-biological peptide monomers, such as augmented proteins from Nanosystems Ch. 15. The downside, is the yields are worse than ribosomes (70 monomers are practical, compared to 300–500 in biological proteins).
People stitch together multiple polypeptides with hierarchical self-assembly to make artificial polypeptide chains longer than 70 monomers (up to 300 may be practical). However, this starts to hit practical limits the longer the chain gets.
Artificial DNA origami regularly reaches much higher monomer counts. The KSRM book reviews multiple projects regarding DNA, including one by William Shih who was already a well-known researcher at the time (~2004).