Leaving aside the financially inspired hype I have observed a strong bandwagon phenomenon that was present during my work in plasma/fusion, higher education, and computer science/artificial intelligence. Popularity outweighed individual critical judgement. It strongly influenced what was publishable and what got funded.
See "bandwagon effect" on wikipedia.
I agree with Linus Torvalds, "I think it's fairly positive that vibe coding means that anybody can get a computer to do something that they otherwise couldn't have done. But from a maintenance standpoint, vibe coding is horrible, horrible. Because you don't understand what you're doing."
For a developer the remaining work of formal specification, testing, and debugging is frequently more difficult than the actual coding, and a lot less fun.
I do use LLMs as a natural language database to retrieve example code. This is as a supplement to my code library.
See also my blog of 20 March 2024.
Singer and Demo argue for* a society of diverse specialist AI agents like I have been building.** Each and every living creature resides in a niche. I believe it will be much the same for mechanical life.
* The Future of AI is Many, Not One, arXiv:2603.29075, 30 March 2026
** See, for example, my blogs of 16 September 2010, 1 February 2024, 20 November 2023, 1 October 2021, etc.
I have previously described* how to give A.s.a. H. AI robots most of the Toki Pona language vocabulary. Sonja Lang's Toki Pona Dictionary** can then be used to chain these core concepts together to build a large concept structure and an English vocabulary of many thousands of words. For example: snow=(cold, moist, white, powder), learn=(information, approaching), brain=(thinking, object), etc., etc.
* See, for example, my blogs of 1 October 2015 and 4 September 2024 and Trans. Kansas Acad. Sci., vol. 120, pg. 108, 2017.
** Tawhid, 2021.
Although I am not a capitalist I would point out that another advantage of specialist A.I.s is the possibility that different specialists might be produced by different manufacturers. Each manufacturer can, itself, be specialized. (This would possibly also reduce the cutthroat hypercapitalist competition between manufacturers.)
Some sensor fusion is done by preprocessors. Conductivity and temperature are used to measure salinity. The product of voltage and current measurement is used to indicate power. Etc.
A good deal of sensor fusion is done by A.s.a. H.'s concept structure* (at various different levels of abstraction).
* as an example see my blog of 4 September 2024.
Be it hardware, software, experimental procedure, etc. you need to provide enough detail that someone else could reproduce your work. Only then can others build upon and advance our science. You will then benefit in turn from THEIR contributions. i.e., Nash's cooperative games.
