Reading about what everyone is actually vibe coding for the last 3 years, it's basically search results from github & stack overflow. They search for BASIC interpreter, weather sensor logger, or something which has been implemented countless times in the past, the AI models are spitting out fragments of a listing from github & stack overflow.
It's very difficult to express much more than a typical search query in natural language or to come up with a truly novel invention. That's why patents are so loaded with meaningless obfuscation & standards are usually written in pseudo code instead of natural language.
Lions once either typed in or ran an AI demo from Compute magazine & initially marveled at its logical reasoning. In natural language, you could tell it A & B was true, B & C was true & it would deduce A & C was true. Lions quickly discovered it used a simple truth table & was best suited to a certain type of logic problem. Then the wonder quickly faded.
The vibe coding phase was the same way. It's now a very large truth table compressed using weights but it's still the same operation.
All record of that program is lost. Archive.org only shows another program called Learner & has no full text search.
The generative art phase had a good thing going. Everything is now generated, but it's too expensive & it looks like rehashed shit. There's still promise of unlimited, decent art while the coding side raises the question of what advantage there is to describing algorithms in natural language instead of a programming language.
Though lions only had a C64 from 1984, a lot of magazines from 1983 must have been loaned from someone else. Lions typed in a lot of listings from 1983. The 1983 issues were quite hefty compared to later issues.
After years of searching for many listings, it appears most are outside the scope of what has been copied to the internet.
Lemonade stand for C64 was actually a demo which came on a tape. Vaguely remember the game but not how it was distributed. Remembered the PETSCII, the word conc. & the brutal weather conditions it printed. Enjoyed it briefly.
Magic Desk was another forgotten program lions remember seeing but never using.
-------------------------------------------------------------------------------------------------------
https://archive.org/details/open-look-the-case-is-clear
Another 90's corporate back rooms video. Open Look came to be used in a lot of early Linux programs. There was probably an open source clone by then. Workman was written in it. Some window managers undoubtedly were. The old man programmed 1 project with it. It made sense that AT&T was paying people to develop on OpenLook, a clone of IBM's long standing OfficeVision program, though it seemed like a badly managed disaster. He truly hated it & suffered anxiety attacks.
AT&T did release an Officevision clone called Rhapsody in 1990, for PC. The Chinese workers of that time were manely displaced by Indians, after becoming too expensive.
You can still write programs using Openlook today, though lions would wonder why. Perhaps the right diction could successfully get a program started. Amazing that a GUI toolkit once formed a significant part of Sun & AT&T's revenue.
The early 90's were a different universe. The Mac GUI had been around for 6 years but the Windows GUI was just getting started, with the world manely using DOS. X11 on super expensive UNIX workstations was being sold as an equal to Windows.
A standard toolkit for UNIX now seems ludicrous. The hodge podge of different toolkits on a standard drawing system is the only way it could have ever been.
The selection is still pretty awful. Lions would say Windows still has the most neutral & useful toolkit but they're all bad. Lions aren't even satisfied with their own toolkit. The best toolkit we ever had was Qt circa 1998.
Comments
Post a Comment