Sociopath Community

Jada said: 

Are you really that much of a pea brain?

If you're not using AI you will get left behind very soon

I think it's good for grunt work that has no quality checks. Like writing a letter to immigration department or something, where you don't want to waste time formatting the letter or making the text formal.

Other than that, AI still has a long way to go before it can do useful things.

"AI still has a long way to go before it can do useful things."

I know this to be incorrect. AI can provide advance code. As long as there is a reference in the AI database. Coding has reference material available in bulk.

Yes it can provide grunt code, like the example you gave to Good.

Jada said: 

Yes it can provide grunt code, like the example you gave to Good.

 Another lame attempt. Try harder.

The whole fad emerging called "vibe coding" seems pretty bogus.  At best it is lazy and unoptimized, etc.  At worst it is reckless and might end up creating more work cleaning it up.  LLMs are another abstraction layer and interface paradigm.  It still has to use existing languages to code with.  LLMs themselves are going to be rate and budget limited, both on the provider and user ends.  They will continue to improve, streamline, and perhaps their continued adoption into the workflow will mitigate those limitations, as they will be seen as a priority and developed with these incentives in mind.  They were limited by just how aware you could make them to the context of your code (over multiple files, etc.), but with stuff like Windsurf and certain extensions, the copilots and such can actually peer into and edit multiple files, getting bigger and bigger windows into context.  Of course, they can't read your mind, so a human is always going to need to be in the loop.  There's no way these things are going to entirely eliminate humans; it will likely change the rates and methods of productivity.  I think it's really too early to say the full impact on coding culture and its workforce.  Most takes are going to be rather naive, because no one really can tell.  Perhaps feed the info into an LLM and ask its predictions?

What might be a more worthwhile endeavor is to look over and optimize the various programming languages already on offer and find a way to either make a universal type, or consolidate the most useful languages per paradigm or use case, as is necessary.  It would help keep the data set for training cleaner and more standardized.  That effort would make the use of LLMs in code a better fit.

LiYang said: 

Good said: 

LiYang said: 

 Your queries are far to general.

You must narrow your queries down to the method or function level. Think more like a software manager instead of a grunt programmer.

junk queries in junk out.

 

for example

Provide a c function that takes an array of 1000 floating point numbers and returns the average of the array.

What you ask is to think like a grunt programmer instead of a software engineer. I cant tell if you are serious or not.

 Are you really that much of a pea brain?

 

If you're not using AI you will get left behind very soon

 When you learn to read and comprehend, I will speak with you again.

To be honest, the gains in automated coding aren't quite as impressive or ominous as the prospects AI agents tease.  Coding is important for AI to master primarily toward the effort of allowing it to self-improve.  An idealized future could see AI and LLM integration as a primary operating system, which could create or composite code on the fly for the needs of the user.  Applications might be more fluid and as-needed, rather than separate and proprietary data.  There might be more prompt sharing and saving instead of precise code, or else a common repository which would streamline the AI's efforts.  This kind of usage might promote a more generalized familiarity with code, as while it might take care of a lot of the heavy lifting, AI would probably still need their human user to customize or tweak things to really satisfy their personal preferences which AI is likely only ever to approximate.

Agentic use will open up even broader use cases.  The limitations with LLMs in the past and copilots now, etc., might all but vanish, since if you are having an agent work on a coding project, it is going to manage the entire file system and its entire code, use GitHub or whatever and create a repository, manage versioning, etc.  They have been able to error detect and correct quite well, but human intervention is currently still pretty necessary in some cases.  Of course, agents would be more than merely coding platforms, although it likes to use code to accomplish a lot of its tasks (Python scripting on the fly to handle data analysis and web-scraping, for instance).

...On the side, I did ask an LLM the best mixture of human languages and programming languages to make token use and prompting and its own "thought process" most efficient (so more cost benefit for the user and concise and information-dense for the LLM):

Claude Sonnet 3.7 said:

Here's an optimized list of human and programming languages that provide high information density while being mindful of token budget:
Human languages:

Mandarin Chinese (character-based writing system packs considerable meaning in few characters)
Japanese (combines kanji, hiragana, and katakana for efficiency)
Korean (hangul alphabet is phonetically precise and compact)
Arabic (root-based system with abjad script conserves space)
Hungarian (agglutinative structure embeds multiple concepts in single words)

Programming languages:

APL (array programming language with unique symbols for complex operations)
J (successor to APL, even more symbol-dense)
K/Q (minimalist syntax for financial/data processing)
Haskell (concise functional programming with powerful abstractions)
Perl (nicknamed "write-only" for its compact but difficult-to-read syntax)

The combination of these languages allows for expressing complex ideas with minimal token usage - human languages that pack meaning into characters/symbols paired with programming languages designed for terseness and dense information encoding.

I also did some querying for an optimal, single programming language.  However, I think it would be a better first stage to try and work out the best features for a particular paradigm first, but that might not be more practical than just trying to get the best out of all the possible worlds right away.

Claude Sonnet 3.7 said:

Building a Unified Language
To create a unified language incorporating these strengths:

Core Foundation:

Start with Rust's ownership model for memory safety
Add ML-family type inference for expressiveness with safety
Include pattern matching as a first-class construct


Modularity:

Support multiple paradigms through flexible syntax and semantics
Allow pure functional code with guarantees when needed
Provide OOP capabilities but discourage inheritance in favor of composition


Concurrency Model:

Adopt Erlang's actor model for distributed systems
Integrate with Rust's ownership for safe concurrent memory access
Include Go-style lightweight threads/coroutines


Developer Experience:

Focus on excellent error messages like Elm/Rust
Build a comprehensive package manager and build system
Design for gradual adoption and interoperability


Implementation Strategy:

Start with an LLVM-based compiler for performance
Design a VM for portability and dynamic features
Create progressive typing system that can be as strict or flexible as needed


Practical Considerations:

Include a foreign function interface for existing code
Provide tooling for gradual migration from other languages
Focus on compile-time guarantees while maintaining runtime flexibility



This unified language would balance the safety of Rust, the expressiveness of functional languages, the practicality of Python, and the concurrency of Go/Erlang while maintaining a coherent design philosophy rather than just being a collection of features.

Honestly I think part of why AI is so overrated is because of the religious unwillingness for society to observe the brain and its functions in relation to environments. Along with anti-marxist anti-materialist constantly having to make sure marxist conclusion are never met with academia even if they have to cripple academia or destroy it.

AI is pretty much the embodiment of capitalist ideology. Stealing from originality and authenticity while lying the entire time about its own existencial properties.
AI also offers no disagreement and no critical thinking. You hear what you want to hear, making you and it both more prone to fallacy.

Personally, I've been quitting AI stuff and deleting my AI images. I've also been hiring real authors and real artists. It's just sad seeing apps with good UI being used for AI instead of adding dynamics to human relations. Maybe the shaming of social media is party to blame, instead of having the idea that social media should be advanced instead of shunned and controlled.

People will hate on phones and social media, but not AI, because it gives them a sense of control.

Buttered Toast said: 

To be honest, the gains in automated coding aren't quite as impressive or ominous as the prospects AI agents tease.

I've used Grok3 to make some very advanced function calls. Verified, tested and in use.

The skill is in designing the query to grok3. Have very specific argument variable and returned variables.

Once again, if you're not using AI, you'll be left behind very soon.

This video is pretty good at showing some example coding.

Good said: 

LiYang said: 

Good said: 

What you ask is to think like a grunt programmer instead of a software engineer. I cant tell if you are serious or not.

 Are you really that much of a pea brain?

 

If you're not using AI you will get left behind very soon

 When you learn to read and comprehend, I will speak with you again.

 Oh noooooo, Good won't speak to me. I'm so sad.

Maybe next time you can have something of value to say.