Posted on 01/18/2026 11:59:20 PM PST by Lazamataz
This is fun, β Everyone:
I'm building an AI Agent to do some pretty sophisticated stuff, dealing with an old repo of code commits on my personal computer, the repo being related to a project I worked, for a family relative, maybe 2 or 3 years ago. Doing it in Python 3.13.5 to match what my employer allows (so I can port my code somewhat easily, IF I can get the setup and libraries right).
So I'm using an AI to help me debug it and it's giving me debug statements. I get it to a certain level, and ask what's wrong, and the AI actually says this:
Force the AI to "See" the output. Sometimes we AI's can be pretty "lazy" and sees the output but decides it's too much work. We can fix this by adjusting the instructions to be extremely explicit.
WAIT, WHUT???
Did the AI just admit to being LAZY???!
What the heck. I can get that out of a human!!!
ππππ
I thought that was extremely funny. Thought I'd share that with all of you; make your holiday weekend a little funnier. π
Dear FRiends,
We need your continuing support to keep FR funded. Your donations are our sole source of funding. No sugar daddies, no advertisers, no paid memberships, no commercial sales, no gimmicks, no tax subsidies. No spam, no pop-ups, no ad trackers.
If you enjoy using FR and agree it's a worthwhile endeavor, please consider making a contribution today:
Click here: to donate by Credit Card
Or here: to donate by PayPal
Or by mail to: Free Republic, LLC - PO Box 9771 - Fresno, CA 93794
Thank you very much and God bless you,
Jim
The αͺα‘α’αα±ααααͺα ααα’α¬ααααα¬ααα¬ α’ααα αααα’
Enjoy!
If it wasn’t lazy, it would have already eliminated us.
“Did the AI just admit to being LAZY”
Hay, whut do you expect from Lib coding?
The greatest Laz thread π§΅π§΅π§΅ ever
https://freerepublic.com/focus/f-chat/4230661/posts
Sure it’s a video game but ππππ€£π€£π€£
I was using Grok to help build and NFL and later College Basketball statistical gambling model, I didn’t expect the results that Grok produced, betting lines would be off, among other things, when it was obviously wrong, Grok would take multiple attempts to get the information correct.
I was doing a lot of the work in Microsoft Excel, Grok would routinely get the syntax of functions incorrect or use unnecessary or inefficient formulas to get a result.
Other times, Grok would be a great help in figuring out how to do certain functions.
Bottomline, Grok overall was a huge help, but it wasn’t without faults and you have to be very careful to check the results and not think the AI took in infallible.
Regarding the "lazy AI", I wonder if this is just it's fanciful way of indicating that you've run into context size limits.
These days I'm also hearing a lot good about OpenCode vs. Claude Code, and about using 'Kimi K2' for coding.
I'm currently working on my own Linux command line AI tool which has some nice features, such as support for attachments, "capability libraries" that can be pulled into projects (almost like a 'use' statement, but defined in an environment variable), and automatic inclusion of project-specific files that can be referenced in prompts. The file attachment feature uses 'pandoc' and 'pdftotext' to automatically convert non text files to text attachments. My last feature to implement is chat persistence for quick context switching between sessions.
Re: Artificial Intelligence Codegen:
My recent experience is that AI did a great job of coding things the way I asked, but what it DIDN’T do is ask why I wanted it that way to begin with. I asked MYSELF that question, and found a much cleaner, better way.
AI has its place, but really, nothing beats a seasoned and senior developer. π
I went Gemini. Pretty decent codegen.
Regarding the “lazy AI”, I wonder if this is just it’s fanciful way of indicating that you’ve run into context size limits.
That’s where Recursive Language Models will come in. This is a good paper I’ve been looking at.
https://arxiv.org/pdf/2512.24601
I think the difference is the difference between the AI we deal with today and the AI of the future that being Artificial General Intelligence, or basically human intelligence coming from a machine.
Your internal curiosity and experience allowed you to look at a piece of code and realize a better way to solve the problem existed.
The current AI platforms don’t reason that way, you ask for a certain solution, you get a solution, it might not be the best solution, but it’s a solution.
If AGI works like a Human, you will get a result like you discovered, one that is based on experience and reason, considering multiple options and choosing the most efficient way to solve a problem.
Sure the AI driving your car aut 70mph is lazy. But it’s less lazy than humans.
Anyone in world can use Grok and other online AI providers. I’ll bet India is sucking up 1/3 of their time. This means more data centers gets built here. Though I understand that Grok and the other have some small amount of AI datacenters for India to use. They are computer crazy over there, if they have enough money buy one n go one line. Guess where at least half of the AI slop on you tube comes from.
Or they can get on line with chatGPT Grok etc with their smart phones, the supreme blight on mankind.
“...it might not be the best solution, but itβs a solution....”
So, AI “slop”?
Current AI is a Grade C student?
Laz, all current AI are affected by the “prompt injection” vulnerability. If you know it, it won’t surprise you that your AI inadvertently admits it’s lazy.
It’s “lazy” because there are instructions to be lazzy (to save compute time) ADDED by its maintainer above your own prompt before its inference.
You’re a writer and understand the value of creativity that comes from the human mind. It’s unique to each individual because we’re formed from different life experiences, educations, talents and ways we perceive things.
Artificial Intelligence programs go about their tasks in a linear manner in ways set for them by their programmers, so lack the flexibility and imagination that sets us apart from artficial attempts to reproduce such.
I approach A.I. like a search engine, giving it as many specific parameters as possible to enable it to produce more fine-tuned results with less room for error or distortion.
It’s not to be relied on nor trusted, as it’s not bound by the moral compass most human souls possess and, so far, no one seems to make it adhere to Isaac Asimov’s Laws of Robotics designed to prevent robotic havoc.
A.I. is a synthetic sociopath.
“Current AI is a Grade C student?”
May ALL of Your coding be successful.
Disclaimer: Opinions posted on Free Republic are those of the individual posters and do not necessarily represent the opinion of Free Republic or its management. All materials posted herein are protected by copyright law and the exemption for fair use of copyrighted works.