Llms are not deterministic and can implement the same spec in diferent ways or slight variations, how can we get deterministic approach so we can gain development speed and accountability over our systems? Can something like that exists ?
Absolute/unchanging clarity of purpose/intent/interaction between various dynamic levels/groups in an ever changing world. (P=NP)
Unified audience and/or obsolescence? airplane demo[0].
Obsolescence lot easier to acheive than 100% unified audience. No effort required for obsolescence.
By definition, system(s) approach or systematic approach? Both provide flexibility ( framework of abstractions ). Flexibility and absolute determanism are opposing concepts. (NP != P)
Programming language(s) & related software environments[1] could be seen as a lot of attempts at reconciling flexibility and absolute determanism. (independent of ever changing intrinsics used over years X86/amd, theory CISC/risc/arm/mips and various programming levels[2])
related concepts: NP != P; C language spec[1] (vs. lisp machine vs. lisp running on conventional von von-neuman hardware )
Theory/what - Unused system(s) haven't been introduced to implimentation variations.
Practice/how - Used Systematic approach(es) will wind up with intrinsic implimentation variations to resolve mismatches between theory/practice.
Each will break/degrade when appropriately used.
Keep it simple, split the difference and use vm/webassembly for testing/debugging. Not very practical, but skip the implimentation & theory can lead to absolute "determinism". Just move on & don't restart after void/segfault clarity happens. It's broke. Don't know why -> clarity without understanding underlying implimentation. What's usually missing is the ability to rapidly move on to something else while the spec always produces the same system state.
[0] : https://meyerweb.com/other/humor/osair.html
[1] : https://en.wikipedia.org/wiki/List_of_programming_languages
[2] : https://medium.com/@hannah.scherz.23/layers-of-abstraction-i...
[3] : https://www.geeksforgeeks.org/c/c-programming-language-stand...
https://hofstadter.io | https://github.com/hofstadter-io/hof
It's built around CUE + text/template (template authoring is the more painful part)
So much for always use lambda expressions in macros and never have to worry about mutex issues.
> What's missing for a true deterministic ....
agreed upon common reference monoid / starting point
> What is stopping us from such language, visualization or construction method?
ADA Spark, Gleam_otp/Elixir/Erlang_otp aren't main stream. All programing languages don't interact natively with BEAM.
Domain/subject area using DAL in/for/with is missing.
OBS studio[0] might be more easy for producing live video streaming/production.
Dynebolic[1] is an OS distribution geared for multimedia production.
vs. DAL's for banking / stock market transactions.
vs. DAL's used in PX4 [2] & ways to add to / expand a DAL [3]
------------------------------------
[0] : https://obsproject.com/
[1] : https://dyne.org/dynebolic/
[2] : https://docs.px4.io/main/en/concept/architecture
[3] : hof : A tool that unifies data models, schemas, code generation, and a task engine., https://github.com/hofstadter-io/hof
---------------------
Beamup with Erlang and Elixir : https://www.youtube.com/watch?v=VIysBF-uWf4&t=203s
The wickedly Awesome Batch and Stream Processing Language You Should Have in Your Toolbox. : https://www.youtube.com/watch?v=4c6tY0dLni4
Gleam_OTP : https://gleam-otp.hexdocs.pm/0.5.0/index.html
Deterministic workflows are more challenging because they are more complex in nature, as soon as you start interacting with the world outside of a pure language itself.
Business view of lims/dal:
DAL (diagnostic assembly line system - compliance) passes things up to LIMS (information management system - accountability/tracking through process life cycle) & ERP (like SAP). Industry/company policies / regulations set what is considered a deterministic resolution for a LIMS/ERP process item aka successful process life cycle completion and/or what needs to be done for a non-deterministic / incomplete life cycle process. Information from or not received from one or more DAL's or other sources may/may not contribue to a non-deterministic outcome/resolution.
Contrived example: "Oh man, music data strucures are a rabbit hole." : https://news.ycombinator.com/item?id=50200065
Project didn't have a deterministic way to figure out if individual songs being added already esisited in 'database'. Song(s) packaged differently so 'same' audio song need multiple methods to figure out if already in database. Not all computation methods / automated methods for finding songs were 'available'. aka no code to determinsticily go through all various ways to find duplicates.
Vs program acoustid provides a unique identifying 'fingerprint' (DAL level item) passed to LIMS -> beets/musicbrains. LIMS then uses unique identifying 'fingerprint' as a way to associate other addition DAL level data (artist, record lable, what else released with). Undiscussed, what if 2 recored lables released same song? LIMS might contain information on which format to keep/convert to if two song fingerprints are the same). Missing information on to ignore file format or convert to a specific file format would lead to a non-deterministic result of only difference between 'identical' song fingerprints was song file format. Different formats (file & recording) do generate different quality recordings. LIMS level usually is level sort/resolve conflicting data from DAL's. Note: 'how to deterministially resolve issue(s)' may be a feature request by a user after LIMS (SAP) has been set up.