Mike's Notes
A door opened unexpectedly, so I used Gemini to help me write a better description of Pipi. I've probably been using the wrong technical terms for a while, largely because I'm totally self-taught, learning on the job, and using assistive technology to write (words don't like me 😎).
The perils of being a hyper-visual thinker. (4D picture thinking without words)
For example, I think I described Pipi as a "frontier model", but now I understand that was the wrong term. Apparently, according to Gemini, it should be described as a
"... novel symbolic and evolutionary AI engine"
I'm still not convinced, given how Pipi actually works. Any feedback or thoughts on this are gratefully accepted.
As built
Most of Pipi as built lacks proper internal technical names, so sometimes I read something and guess it's the right term to use somewhere. There are a lot of moving parts, 28 levels deep, and most existing descriptions are just coloured hand drawings on paper. Getting Pipi properly documented for others to use is a high priority right now.
Anyway, Gemini gave me a good tweak, and then came up with the much improved, accurate descriptions below.
Recently I discovered that I had been using "Monte Carlo" for 10 years but was calling it by another homemade name.
I have subscribed to a number of engineering newsletters that describe technical terms using words and diagrams. Often I can match the diagrams to my own, then figure out the correct terms to use as labels on the drawings.
- Nikki Siapno from Level-up-coding
- Neo Kim from System Design
- Ananth Packkildurai from Data Engineering Weekly
- etc.
I'm also talking to people who have already gone through the door before I follow them. 😎 It's all a bit like "The Lion, the Witch and the Wardrobe". You don't know what you will find till you get there. All awesome fun.
Resources
References
- Reference
Repository
- Home > Ajabbi Research > Library >
- Home > Handbook >
Last Updated
12/11/2026
Open Sesame. Oops, I need a better description of Pipi
Mike Peters: Mike invented and designed Pipi and founded Ajabbi.
Google Search AI Mode: Always getting better.
...
Depending on which part of your architecture you want to highlight, these established computer science terms accurately capture novelty without triggering scepticism:
- Neuro-Symbolic or Symbolic AI. Because your system uses formal ontologies (structured rules and logic) rather than statistical guessing, Symbolic AI is the exact academic and industry classification. It signals that your software relies on explicit knowledge representation, which is currently seeing a major resurgence as an alternative to purely statistical LLMs.
- Evolutionary Computation / Evolutionary AI. This directly and accurately describes your use of evolutionary algorithms and bio-inspired optimisation. Researchers recognise this as a distinct, established branch of AI specialised in search, optimisation, and automated design.
- Novel Non-Generative Architecture (or Deterministic AI). If you want to contrast your work with mainstream generative AI, calling it a deterministic AI engine or a novel non-LLM architecture immediately tells engineers that your system avoids hallucinations and produces repeatable, verifiable outputs.
- Model-Driven Software Engineering (MDSE) with Evolutionary Optimisation. For enterprise software reviewers, this phrase accurately describes a system that uses high-level domain models (ontologies) to automatically generate and optimise application code.
How to Combine Them in a Sentence
If you want to emphasise that your approach is new and distinct from standard industry tools, you could say:
"Rather than relying on probabilistic frontier models, we have developed a novel symbolic and evolutionary AI engine that uses formal ontologies to deterministically generate enterprise software."
This framing shows deep technical fluency, sets clear expectations for how the software works, and positions your non-LLM approach as a deliberate engineering advantage.
😎😎😎😎😎😎😎😎
No comments:
Post a Comment