Pipi 8 (2021-2022)

The microservice architecture wouldn't work because the simulator induced exponential complexity. It was going to be a big scaling problem. I had a good think and decided to go into the deep end. Instead of trying to reduce complexity - why not maximise it and then harness it. Just like in nature.

2021 Start

The key to the puzzle was introducing global namespaces and changing the architecture so any messaging could be 100% reliable.

For security, I tried deeply nested microservices. So good so far.

Then, I introduced Algorithms and Markov Chains, fuzzy logic and noise injection.

PostgreSQL became the production database. Thousands of parameters were needed.

In all this, the self-documentation broke, so I was working completely blind. Pipi was headless, and there was still no API, so I only knew it was working by looking at the logs.

Rich primitives based on the objective reality of emergent layers replaced some entities. Each was a software program and ultimately derived from the basic laws of physics and replicable science.

I finished up before Christmas.


Wikipedia became my friend. I had to take a deep dive into algorithms. And it is amazing what you can find by googling. Every day I was often watching a conference talk from QCON.

But Wait There's More

It worked but no one could use it. The headless state of Pipi needed fixing.

Markus Covert and Mycoplasma

Mike Notes

In 2014, I read an article in Scientific American by Markus Covert at Stanford and the successful computer cellular simulation of mycoplasma. It was a fascinating article that made me think about PIPI and whether it could have been built better in 2005-2008.

When my wife and I left Christchurch in 2014 to start a new life, I became curious about modern cloud computing and read a great deal. I then realised that PIPI was a very early form of cloud computing, and many of its features were 10 years ahead. What a waste!

By 2016, I got wild about losing the opportunity and decided to rebuild PIPI as a core platform from memory.

From 2017 to 2019, PIPI 6 was built by fusing the rebuilt core platform with Covert Lab's open-source cellular simulation software. It was challenging to do and was the cause of an unusual architecture.

Every process was given a random probability. And a lot of positive and negative feedback loops. It was a bit like composing music (In hindsight, synesthesia helped a lot). 

A Computational Whole-Cell Model Predicts Genotype From Phenotype

A 2013 talk by Markus Covert at the Qualcomm Institute.