IDE + AI : IDE = IDE : Notepad (Follow-up to "AI or DIE")

[ASBzone]

One thing I noticed is that in addition to artificial intelligence it seems to have acquired artificial laziness because in one example where it had to translate about a dozen printf(s), it translated 2 of them then put a comment { add more here... } ...  I got a good laugh out of that (who's working for whom here ???.)     Its artificial laziness is pretty darn close to the real thing.

Oh, the laziness is real...      What AI exposes is that humanity still does not have a firm grasp on how humans use their brains, and how their brains use them (i.e. procrastination, laziness, etc)

The AI never agreed to work for you personally...  Just remember that.    It's don't gig work and helping us out every now and then.

[schuler]

@440bx,
Very sorry for the late reply. I am already snowed under with my current projects and I can not absorb more work ATM.

@all,
Just had a new record with Claude Sonnet 4.5 + https://github.com/joaopauloschuler/beyond-python-smolagents: 3293 lines of code with a single prompt.

This is the result: https://github.com/joaopauloschuler/ai-coding-examples/blob/main/src/free-pascal/task-manager/task-manager-claude-sonnet-4.5-20251006.pas

The pascal source code was produced at the third iteration of "evolutive_problem_solver". It took about 2 hours to run and then it crashed into my spending limit: US$ 8 for this experiment.

This is the code used for the testing (I changed the model before running): https://colab.research.google.com/github/joaopauloschuler/beyond-python-smolagents/blob/v1.21-bp/bp-examples/pascal/pascal-single-file-01.ipynb

[440bx]

@440bx,
Very sorry for the late reply. I am already snowed under with my current projects and I can not absorb more work ATM.

No problem.  That's life.

[schuler]

Just had a new record with Claude Sonnet 4.5 + https://github.com/joaopauloschuler/beyond-python-smolagents: 3293 lines of code with a single prompt.

This is the result: https://github.com/joaopauloschuler/ai-coding-examples/blob/main/src/free-pascal/task-manager/task-manager-claude-sonnet-4.5-20251006.pas

The pascal source code was produced at the third iteration of "evolutive_problem_solver". It took about 2 hours to run and then it crashed into my spending limit: US$ 8 for this experiment.

Just to comment that I've just done a similar experiment this time using evolutive_problem_solver_folder instead of evolutive_problem_solver. It costed me about US$10 with API calls. It produced about the same number of pascal lines. BUT, it also produced about 1500 lines in md files... It is not perfect. But it is interesting...

This is the result:
https://github.com/joaopauloschuler/ai-coding-examples/tree/main/src/free-pascal/task-manager/task-manager-claude-sonnet-4.5-20251021

Long file to pascal + AI.

[schuler]

Hello Friends
Just 5 weeks ago, I was creating 4500-5000 lines of code (3000 pascal + 1500 md) with a single prompt.

I HAVE NEWS

The AI script produced about 18000 (9K pascal + 9K md) lines of code with a single prompt. This is the code produced by AI:
https://github.com/schuler-experiments/solution1/tree/solution3

In the "solution3" branch above, all commits were AI generated.

This is the script that produced 18000 lines with one prompt:
https://github.com/joaopauloschuler/beyond-python-smolagents/blob/development6/bp-examples/scripts/run-agent-cycles.py

• model: (default) Claude-Haiku-4.5.
• cycles: 20.
• steps per cycle: (default) 100.
• prompt: (default - press ENTER).
• planning interval: (default) 22.

It crashed after 2 hours running. The bug has been fixed. It would probably go beyond 18000 lines.
COST IN API CALLS: about US$4.

Long file to pascal + AI

[creaothceann]

Just 5 weeks ago, I was creating 4500-5000 lines of code (3000 pascal + 1500 md) with a single prompt.

I HAVE NEWS

The AI script produced about 18000 (9K pascal + 9K md) lines of code with a single prompt.

Long file [sic] to pascal + AI

Measuring programming progress by lines of code is like measuring aircraft building progress by weight.

The statement, "Measuring programming progress by lines of code is like measuring aircraft building progress by weight," is an effective analogy highlighting that both metrics are superficial and misleading proxies for actual progress or quality. Both are easily manipulated and fail to capture the true complexity, efficiency, or value of the work being done.

Why Lines of Code (LOC) is Misleading

• Quality over Quantity: A few lines of well-written, efficient, and maintainable code are vastly superior to hundreds of lines of buggy, convoluted, or redundant code. Good programming often involves reducing lines of code by refactoring and creating elegant solutions.
  
• Ignores Complexity and Value: A simple program to print "Hello World" can take a few lines, while a complex algorithm for medical diagnostics might also take a modest number of lines but represents far more effort and value. The LOC metric ignores the intellectual effort required.
  
• Encourages Bad Practices: This metric incentivizes developers to write verbose, inefficient code and to copy-paste existing code rather than creating reusable functions or libraries, which ultimately slows down progress in the long run.

Why Measuring Aircraft by Weight is Misleading

• Efficiency and Design: Modern aircraft design prioritizes strength and safety while minimizing weight to maximize fuel efficiency and payload capacity. Adding weight arbitrarily for the sake of a metric would create a dangerous, inefficient, and possibly non-functional aircraft.
  
• Focus on Functionality and Safety: The primary measures of aircraft building progress are structural integrity, adherence to strict safety regulations, functional systems testing, and flight readiness. Weight is merely a design constraint, not a measure of success.
  
• Value of Engineering: A completed, flight-certified lightweight aircraft is the goal. A partially completed, excessively heavy structure is a failure, regardless of its total weight.

Conclusion

In both scenarios, the true measure of progress is the successful implementation of functionality, quality, efficiency, and meeting specific engineering requirements. Using a simple, quantifiable, but ultimately irrelevant metric like LOC or weight fundamentally misunderstands the nature of skilled work.

[cdbc]

Hi
@creaothceann: Whoah - You're barking up the completely wrong tree here!
I'm with Dr. Schuler on this one =^
@Joao: Nice work mate  Very impressive.
Regards Benny

[Thaddy]

@schuler
I take it that in your case the $4,00 is just the  latte macchiato?
Impressive results.

Good to see that your own field of research lead to such fine results.
Good example on how to use it too, not only develop the models.

[Lauriet]

Whats wrong with NotePad and a compiler ????