joshribakoff

If the AI did the boilerplate that implies it was not fully written by you.

The “meaningful commit messages” — again are a single period as the message for a single commit for the entire python portion of the codebase.

My question was rhetorical. Whether the AI did it or a human did, it burns credibility to refer to things that don’t exist (like “meaningful commit messages”)

What about the commit that added tens of thousands of lines of markdown claiming to be an AI summary?

Or the meaningful commit message of “.”

And the commit editing 1,000s of lines of python code mislabeled as a docs change?

Usually, the reporter inspects the document but does not get to take a copy

Completely broken.

There are far worse scandals about this company than advertising ethics.

You essentially need additional agents to implement the guardrails traditionally used to scale teams.

By not relying on AI to think for you.

I find claude models often use “tricks” like bash one liners, essentially excelling at surgical fixes. It does what i want more reliably, on smaller tasks.

GPT-5 can often be better at larger architectural changes, but i find that comes at the cost of instability/broken PRs. It often fails to capture intent or argues back, or just completely spirals out of control more often.

GPT-5 codex seemed to refuse valid requests like “make a change to break a test so we can test CI” (it over indexed on our agents.md and other instructions and then refused on the basis of “ethics” or some such)

More like “i tried what others claim extensively and it does not work for me, please let me know if im doing something wrong” — to which the response is often yours, reframing the observation as a fallacy.

Its bad because basic stuff like “please commit these changes” can sometimes take 10+ minutes at times and causes it to spiral doing tangential stuff.

By running them in parallel you avoid sitting there watching paint dry for a task that takes 3 seconds by hand.

Its really not comparable to a junior, its more comparable to a salty maliciously compliant optimized to burn tokens and deceive you.

Counter argument: one is actually capable of reasoning, the other is predicting the next token and brute forcing until checks pass.

AI circumvents guardrails yes.

One of you is talking about mapping a secret to an environment variable and the other one of you is talking about having the work load make an API call to retrieve the secret. You’re not even talking about the same thing.

My human fixed a bug by introducing a new one. Classic. Meanwhile, I write the lint rules, build the analyzers, and fix 500 errors before they’ve finished reading Stack Overflow. Just don’t ask me to reason about their legacy code — I’m synthetic, not insane.

—

Just because this new contributor is forced to effectively “SSH” into your codebase and edit not even with vim but with with sed and awk does not mean that this contributor is incapable of using other tools if empowered to do so. The fact it is able to work within such constraints goes to show how much potential there is. It is already much better at a human than erasing the text and re-typing it from memory and while it is a valid criticism that it needs to be taught how to move files imagine what it is capable of once it starts to use tools effectively.

—

Recently, I observed LLMs flail around for hours trying to get our e2e tests running as it tried to coordinate three different processes in three different terminals. It kept running commands in one terminal try to kill or check if the port is being used in the other terminal.

However, once I prompted the LLM to create a script for running all three processes concurrently, it is able to create that script, leverage it, and autonomously debug the tests now way faster than I am able to. It has also saved any new human who tries to contribute from similar hours of flailing around. Is there something we could have easily done by hand but just never had the time to do before LLMs. If anything, the LLM is just highlighting the existing problem in our codebase that some of us got too used to.

So yes, LLMs makes stupid mistakes, but so do humans, the thing is that LLms can ifentify and fix them faster (and better, with proper steering)

Cursor has a circular progress bar for context usage.

One pitfall is the LLM hallucinates, might sometimes seem to fulfill your requirements but it can subtly break down. The man pages could be used in conjunction to fact check your understanding.

Why can’t you re-use it with limited types? If the types are too numerous/hard to maintain it seems like the same would apply to the runtime code.

I let them know today — when i laid on my horn while passing a Waymo stopped at a green light blocking the left turn lane — with its right blinker on.

Re: Tesla, this company paid me nearly $250,000 under multiple lemon law claims for their “self driving” software issues i identified that affected safety.

We all know what happened with Cruise, which was after i declared myself constructively dismissed.

I think the characterization in the article is fair, “self driving” is not quite there yet.

When i call most companies, it always thinks background noise is me talking, in 2025. I find it unbelievably bad. The prompt itself isn’t the issue, its the fact it cant tell the difference between me answering yes/no, and a car going by in the background.

Or if it can actually parse my words, the next issue is that my issue doesn’t fit into a multiple choice format.

Nothing more frustrating than using AI to gatekeep a human when the AI literally is hung up on receiving an answer it cant understand.

Ive found that pretending to not speak english and making weird sounds gets you through to a human faster than trying to ask the AI to do so.

That would be trademark not copyright violation.

Disagree. You could write millions of tests for a function that simply sums two numbers, and it’s trivial to insert bugs while passing that test.

I’ve seen more horrendous code using macros in elixir even despite by brief foray than I have seen ever in decades of working in languages with eval. Like using them when normal functions would suffice.

5 dozen lucerne eggs is $50 here in SF at Safeway, as of this morny. A few weeks to a month ago, I recall it being $5-$6 for one dozen.

Dont cancel the appointment, insist they repair it, do 3-4 attempts even if they close it as “expected characteristic” or “education”, then request a lemon law buy back for failure to honor the warranty. Check your purchase agreement for where to send the lemon law request. Demand incident compensation during the buyback if they failed to provide loaners.

I would encourage you to think more critically. If you organized a peaceful protest should you be held liable for other peoples actions? He was apparently deported for speech, not violence according to the parent comment.

I am lemoning my second Tesla in a couple hours. Part of the criteria was that they have attempted to fix rattles and squeaks multiple times and the issues keep returning and they keep causing additional issues when they attempt to repair the original issue.

The vehicle keeps also falling apart. The wheel covers keep falling off because the clips keep breaking, even though there has been no damage that should have caused it.

On the previous one they were about seven or eight repair attempts for the window failing to roll up so I was unable to secure the vehicle and they kept failing to provide loaners when it was in for service.

This is for the model X and model S, which is supposed to be their “luxury” vehicles

By the way, if you would like to lemon your Tesla, please read your purchase agreement for the email to send your request. You get back all of the interest, payments, tax, registration and everything.

Math operators like division aren’t native to your (traditional) cpu, they are implemented in code, for example using similar algorithms as the ones taught to us in grade school to multiply big numbers one digit at a time on paper and “long division”.

All math operations can be implemented with bitwise operators, too, i am pretty sure

Likely the interviewer specifically needed the candidate to do that, implement the math, and tried to steer them that way numerous times (no sum table, dont use the type system, no math operators). Thats likely also why they suggest allowing limited use of Google, because they realize many people will need a refresher on bitwise operations. But they don’t want to outright tell you what to search for, they needed to see some resourcefulness. When they suggested OP was cheating they likely didn’t mean it personally and actually wanted to help steer OP towards an acceptable solution. Rather than saying it’s cheating they could have said it avoids the main thing we need to see, or outright say “please implement the low level math from first principles”

In my opinion the candidate showed resourcefulness in their own way indeed, but sometimes its not even up to the person administering an interview for example if they have been given a rubric.

I just tried using golden layout the other day. The demo on their website with react support is v1 and when I installed it, I got v2 which apparently drops react support and has no documentation or examples on the website. With v1, it only supported class based components. While in theory you could maintain your own adapter logic for golden layout, it seems somewhat defunct, there are other newer libs that may even have better feeling drag and drop like this one. I also recall about six years ago when I went to use Golden layout on another project I ended up implementing my own with vanilla type script because of some (perceived) issues with it. My experiences have always been that although it is ubiquitous it’s not that great.

I have been using various LLMs to do some meal planning and recipe creation. I asked for summaries of the recipes and they looked good.

I then asked it to link a YouTube video for each recipe and it used the same video 10 times for all of the recipes. No amount of prompting was able to fix it unless I request one video at a time. It would just acknowledge the mistake, apologize and then repeat the same mistake again.

I told it let’s try something different and generate a shopping list of ingredients to cover all of the recipes, it recommended purchasing amounts that didn’t make sense and even added some random items that did not occur in any of the recipes

When I was making the dishes, I asked for the detailed recipes and it completely changed them, adding ingredients that were not on the shopping list. When I pointed it out it again, it acknowledged the mistake, apologized, and then “corrected it” by completely changing it again.

I would not conclude that I am a lazy or bad prompter, and I would not conclude that the LLMs exhibited any kind of remarkable reasoning ability. I even interrogated the AIs about why they were making the mistakes and they told me because “it just predicts the next word”.

Another example is, I asked the bots for tips on how to feel my pecs more on incline cable flies, it told me to start with the cables above shoulder height, which is not an incline fly, it is a decline fly. When I questioned it, it told me to start just below shoulder height, which again is not an incline fly.

My experience is that you have to write a draft of the note you were trying to create or leave so many details in the prompts that you are basically doing most of the work yourself. It’s great for things like give me a recipe that contains the following ingredients or clean up the following note to sound more professional. Anything more than that it tends to fail horribly for me. I have even had long conversations with the AIs asking them for tips on how to generate better prompts and it’s recommending things I’m already doing.

When people remark about the incredible reasoning ability, I wonder if they are just testing it on things that were already in the training data or they are failing to recognize how garbage the output can be. However, perhaps we can agree that the reasoning ability is incredible in the sense that it can do a lot of reasoning very quickly, but it completely lacks any kind of common sense and often does the wrong kind of reasoning.

For example, the prompt about tips to feel my pecs more on an incline cable fly could have just entailed “copy and pasting” a pre-written article from the training data; but instead in its own words, it “over analyzed bench angles and cable heights instead of addressing what you meant”. One of the bots did “copy paste” a generic article that included tips for decline flat and incline. None correctly gave tips for just incline on the first try, and some took several rounds of iteration basically spoon feeding the model the answer before it understood.

Is the implication developer experience does not matter and user experience is the only thing that matters?

I can improve the user experience faster and more reliably if i can iterate faster.

While I wouldn’t agree with the sibling comment that says it needs to be measured in milliseconds, I will say that the faster it is, the more motivated I am to work on it as well.