since the end of the push back season, i’ve seen an exponential amount of vibe coded projects published both in this forum and in many other places, including entire “libraries” of goop
some of the stuff that’s been made has been pretty good and still seems to have had a certain amount of effort put into it, while a lot of it can really just be considered slop.
what are yall’s opinions on this and how it affects the vex community as a whole?
Imo AI is a tool. It should totally be utilized, but not to write the entirety of a robot’s code
I think the community, and society more broadly, is going to have to learn that there’s a big difference between vibe coding something for your own use and actually engineering a product for other people to depend on.
There’s a lot of slop right now because tools like Claude Code have made people believe they can skip learning the basics of software engineering and architecture. Getting something to work is one thing. Making it secure, performant, private, scalable, and maintainable is something else entirely.
That doesn’t mean vibe coding is bad. It’s great for experimenting, learning, and building tools for yourself, and some genuinely good projects have come out of it. The problem is when someone takes an untested prototype and presents it as a reliable library or finished product.
I think this will sort itself out over time. People will learn what is actually valuable and what only looks impressive at first glance.
For VEX, the main principle should still be that the code a team runs needs to reflect the team’s skill level and understanding. If you use an AI-generated or poorly structured library, you should still be able to explain how it works, troubleshoot it when it breaks, and take responsibility for what it does. AI can help you write code faster, but it shouldn’t replace actually understanding the code.
I have also noticed that, and it honestly depresses me. I haven’t been in the V5 scene for long (I formed 53787N at the end of High Stakes), but my family has been around since ITZ in 202Z. Looking back at those days, I think it’s honestly degraded between now and then. I agree with @Chickee partially when he says this:
Yes, AI is a tool, however that doesn’t mean we should use it. I have used GPT once to make a practice schedule for our team based on availability of members, but other than that, we have a strict no-AI policy on our team.
In short, AI is acceptable for:
AI is not acceptable for:
Past GM issues, AI is also trained on certain proprietary libraries. Using AI comes with the risk of infringing copyright or committing plagiarism.
Always, and I mean always, check your libraries to make sure they’re human-made before continuing to use them.
If you use libraries and AI, good judges and JA’s might sniff it out. AI is only a tool for when you are stuck and no one can help. I would avoid it as anything more than a help.
As long as it is universally available to every student, it is a fair game.
Students will experiment with it, learn how to use it to the full potential, or realize that it is easier to learn programming themselves rather than keep explaining to the AI agent that robot didn’t pick that game piece correctly and it needs to change the code, but it is impossible to describe how.
Btw, an AI agent could be a very patient mentor to teach students how to code. I would encorage everyone to ask their AI agent to teach them “how to fish” instead of “catching a fish” for them to eat.
TL;DR at end.
IMHO AI is a tool that can and should be used, especially for coding. However, what sets a team that uses AI as a tool and uses AI as a crutch is their depth of fine tuning and their ability to explain their code in their Notebook and to the Judges.
A relevant anecdote is when my team member tried to use AI to code our robot, but then it ended up with 23 compilation errors. When he finally got it working, it was poorly optimised, and it took him way longer than it should have. On the other hand, I treated AI as a dev who didn’t really know how to think for himself, and I fed it all the syntax, and my flowchart, algorithms and ideations. The AI didn’t really know what was going on, because VEX especially is such a niche topic, but I fully knew what it was doing, simply because I had told it granularly what to do.
This meant that the AI didn’t really make any mistakes, nor was it slop, and the being which made mistakes/bugs in the code was me. This meant that I could also easily fix the code. Because I architected the algorithm and just made AI do the dirty work of typing out the program (I even debugged it myself lol), it meant that I was fully confident in how each and every line of the code worked and I could explain it in great ddetail in the Notebook and to the Judges.
I firmly believe that AI must be used in this way, because then otherwise it makes too many mistakes and is actually really inefficient…
TL;DR: People should use AI extensively in VEX/Coding, but they should be the architects of the code and design the algorithm themselves to prevent slop. AI will never make a bad coder good, it will make a good coder 20x faster at writing code at the same quality they usually do.
I fully agree. I didn’t see your comment before I wrote my own would have saved me a lot of trouble lol. I always mention a personal anecdote in regards to this topic of discussion…
AI will never be able to make a bad coder good. AI will only ever make a good coder write code to the standard they usually write code at, but around 20x faster.
Story time 
In Push Back, my teammate claimed that he could code the entire robot/auton using Claude Code in approx. 20 minutes. When he did, he never actually compiled or tested it, and told all of us (in a very self-assured manner) that he had fully finished the auton code.
Meanwhile, I (lead programmer) architected the program and wrote the algorithm myself. I then used AI to only write the code, and never really left it to make any descisions on its own. Prettty much all I used Claude for was to save me the wrist pain of typing for 5 hours straight (although the algorithm was pretty long so i did that anyways lol).
My teammate’s code has 23 compilation errors and failed at Nationals.
My code executed decently well, and I was able to EXPLAIN how my code worked to the Judges.
(For context, this all happened the night before Nationals, and I stayed up till 5 am coding)
We won Think award, even though we placed dead last in the tournament.
PS. here is the code which got us Think Award at Nats and qualified us to Worlds:
I agree with that to some extent. However, my concern is that many individuals are using these tools, as they’ve recently grown powerful enough to architect entire projects on their own.
My greatest concern is that much of the strategy and “innovation” components of VEX has been delegated to these AIs. Sure, these programmers can explain it during the interview as they do in fact have a rough understanding of how it works that the tool gave them, but they didn’t do any of the work implementing or deriving these strategies out.
This almost entirely ruins the advantage that those creative programmers have which I think is a very sad state for the VEX community to be in.
I see your point and largely agree. There will always be people who use AI tools the right way: to learn, experiment, and improve. But there will also be those who let the tools do all the work for them.
To be honest, programming a VEX robot is usually not that deep. Most code stays at a fairly basic, entry-level level. The most sophisticated parts are typically tasks or a few advanced features, with the exception of the small group of people who built PROS, JAR, EZ-Template, and similar frameworks.
A strong programmer using AI will almost always outperform a weak programmer using AI. That said, even a weaker programmer who knows how to ask the right questions can now generate competitive code.
One of the biggest issues is that competition judges are rarely equipped to properly evaluate code quality. An experienced software developer can usually spot whether someone truly understands their own code in under 30 seconds. Most judges simply can’t make that distinction.
The reality is that AI is here to stay. It will help us build incredible things, but it will also create real challenges around integrity and learning. VEX rules are clear: you’re expected to write your own code and fully understand it. Beyond that, it comes down to personal integrity.
Outside of VEX, anyone aspiring to be an engineer needs to become very proficient with AI tools. I doubt there will be many engineering jobs in the next decade that don’t treat strong AI fluency as a basic requirement. The real world already operates differently from education and competition settings.
Okay, I feel like I’ve needed to say something about this topic for a while.
I used AI quite a bit last year to help research mechanism designs (not robot design. ex. linkges that resulted in most’ly linear movement) and to help me write better code. Previously, all of my coding knowledge was self taught, meaning I did alot of stupid things, and I had a few knowledge gaps (some OOP stuff, pointers, and structures). However, AI never had the final outcome in my program. I used it purly as a mentor to increase my knowledge.
At one point, I wanted to see AI’s capabilities at writing full files of code (just for fun. I had no intention of using this for a competition at all). Amazingly, after inputting the structure I wanted (using something similar to a state machine to easilly access controller inputs and sensor data, having a basic autonomous selector, and some basic auton stuff like PIDs), Chat-GPT gave me 3 files that actually worked perfectly together.
I feel like the idea that AI can’t write code for hyper-specific enviorments like Vexcode or PROS is hevily exaggerated. If you give it the docmentation, it can easily understand th correct syntax for everything and produce a reletively finished product.
Like how others said, this is now a moral issue, but it is one with some interesting aspects to it. Cheaters are always going to cheat, and you can’t reasonably stop them. However, I beleive that AI has it’s spot in STEM. Just a few days ago, I heard that AI has solved some of the more well-known Erdos problems (unsolved math problems). AI is powerful enough to complete many differrent tasks it is given, and now we as a community need to decide what we will do with this tool (although the line between “tool” and “human replacement” is becoming less and less defined).
I beleive that AI should be taught to students as a tool to use. If we know how to use a tool propperly, we will be better and more efficient at our task. When you don’t teach students how to use this new tool, that is where you get the issues. There are very few limits to what AI can do, and no real regulations given to these kids who haven’t really developed a sense of morals yet.
I guess the old phrase “Give a man a fish, feed him for a day. Teach a man to fish, feed him for a lifetime.” makes sense here. Here we have the option to just take the fish AI might be able to give, even though it’s a small trout, or learn how to fish with AI, catching tuna, mahi-mahi, and salmon.
You are spot on. New AI tools (models and agents) are exceptionally good at working with hyper-specific environments when you know how to use them and provide the right context.
Schools should integrate AI tools into their curricula, but right now most institutions are in panic mode. As a result, students are largely left to teach themselves proper usage. Most colleges face the same issue, with faculty opinions on AI all over the map.
As I have one senior in Purdue’s engineering program, I would caution on the use of AI. In all of my student’s course work and internships, AI did not accurately lead to any quality decisions. Perhaps the questions put to AI were not as good as needed, but programmers who relied on AI or source codes have struggled at the next level. AI is only a tool. Knowledge is power. If AI can do the job…you are no longer needed so why should you be getting paid? I coach my judges on things to ask about coding, while we cannot prove AI was used. The failures to explain things in the interview are taken into account and will impact the results. Perhaps more JAs should step up their game.
I know some colleges have found ways to avoid AI being used to create code or such items. AI is only a tool. Schools don’t teach how to use a hammer or a screw driver. AI is only a tool like a screw driver. I wish VEX would just say we are running things through these AI checkers. The teams will check themselves to be able to compete. If you set the bar at say 85% AI used to create wither code or the notebooks, the teams will check themselves. We can have the teams have a discussion with a team of people if they feel their team is getting flagged when it should not be flagged. Same is done in college. If AI can do the job, the job is no longer needed.
In my opinion, AI use is never acceptable in VEX (or really any context). Ai is not a tool, it just does things for you. A wrench is a tool, a plumber that I hire is not.
I personally agree. I also do believe that there is a meaningful difference between what a decent programmer chooses and what you see AI select in the design of a program. However, these differences only become obvious upon code inspection. Unfortunately or fortunately, depending on who you are, VEX isn’t a software design competition, and thus there is no obvious meaningful difference on the performance side between a student who is good at programming and a student who is alright and uses AI for decisions. My concern really lies on the difference between those who are “good at prompting” and those who are good at software engineering/design. I do believe that being a good prompter is important in the industry, but to do so you need to have a stronger baseline of judgement than the AI, which is what these prompt-maxxing individuals lack compared to those who write their own code, with perhaps some checking done by AI.
This is where the interviews are critical. My compettions use 2 interviews at a minimum to get an award. So hopefully each team sees no less than 4 judges. All the judges communicate during judging so times are held constant and sometimes follow up questions come from the other judges. Teams that generally use AI have problems under questioning. If a team is caught using AI, that is a DQ level issue. I have had some that we had a strong feeling about, but not enough to DQ them. I hope to never DQ a team through the process, but it is up to the teams to play it straight. Each team is accountable for their actions and decisions. I hope this helps some. Too many JA’s undervalue the interview process. I may push the envelop in the other direction. If you win at the events I am the JA at…The team of judges for that event have determined you were the best that day. I will give it to the teams…Lately it has been a harder and longer debate.
I think AI should only be used to review code and recommend solutions, it should never write all of it.
This is something we really do need to talk about. Over the past months I’ve been reading and learning a lot about AI in general and I think I can add a little to this discussion.
First off, I would like to quote some things that others have said and that I believe are true and worth restating
Yes this is so true. The fact of the matter really is that AI is incredibly good at coding and is able to do much more than a lot of people realize.
This is put really well and if you only take away one thing from this topic make it this paragraph.
@Jedi_53787N really does a good job of laying out what is and isn’t legal in VEX. I would quote it but it’s quite long.
The amount AI slop is definitely increasing in this community and others and I fear it will only continue to increase. These agents are getting so much smarter I’m going to assume most of you know about Fable 5 and Kimi K3 they are so smart and it just so happens programming is where they excel.
I use AI in pretty strict moderation these days limiting it to an advanced google search and grammar checker. I do this because I want to learn software engineering and having an agent do all the coding for you is not the way (I do have AI teach me a good bit though). This is where I think there is a legitimate problem skipping the learning proccess and just having it done for you is so easy now and very accessible. When it is not a necessity to learn to get the final product unfortunately most just don’t learn. One quick note, at this point in time pure vibecoding eventually breaks down when you scale.
We should be very careful when using vibecoding and in VEX I believe we should avoid it all together in the robot code and whatever else we make. Many of us will go on to have careers in STEM and it is so important to learn the foundations that are needed. Having AI do things for you will not lay the foundations you need. Vibecoding is not inherently bad but it will not teach you anything except prompt engineering.
I understand that people take the view that AI such as Claude Code is “just a tool”, however it is a tool that can do a significant amount of work, not just for programming but the thinking skills in building a system. Using it without learning how to use it best reduces programming to simply telling the “tool” to do what you want. There is no skill in that. I recommend before pursuing ai based programming for your team to check out the Google Ai Professional course, its completely free and would mean the use of ai would actually teach your something. Use vex as an opportunity to learn, not to delegate what is actually a very fun and exciting (yet sometimes frustrating 
) area of work.