Oh man, I asked Claude to summarize my commits and changes today, using the voice and structure of my usual daily updates.

It was fucking awful.

Claude spent 5 minutes on this then I had to review and edit anyway. I ended up spending 10 minutes writing down the what+why in a bullet list from scratch and it’s 100x clearer and better.

The AI neither did a better job nor saved me time.

I think you’re doing the right thing.

I suspect in five years or so final exams will be a 1 on 1, face to face meeting with the teacher where they ask you some softball questions you should be able to answer easily if you did even some of the homework.

I'm in a field that decidedly does not value the written word beyond the minimum they need to understand and write emails, but I had to take a couple of writing classes in college.

I had to write several 10+ page papers, including research papers (like actual research, not "do your own research" research) before LLMs existed and I'm so fucking thankful for it. Nearly every single day I have to read the shit my colleagues write, and it's like they've never read a book in their entire lives. It's honestly tragic.

And this is a field full of ostensibly intelligent people.

I love this take. One suggestion I would make is also rake your knowledge and try to help someone earlier in their journey. Writing is great but you get no feedback immediately(you may get some later).

When you explain something which you only have a surface knowledge of and the other person asks why... That's teaches both of you alot.

Watching so many functioning adults and people alike just lobotomize themselves immediately when AI became a thing is just so fucking tragic

They’re probably voting though…

People used to call me the dictionary for having a ton of useless information, but now they call me chatgpt :(

I genuinely worked hard on learning an encyclopedia worth of knowledge and fun facts and I get compared to a hallucinating predictive text model. SMHOAT

The longer comments I write get the fewest upvotes.

That's crazy—It seems implausible to do that. As far as it goes for me—I can't imagine writing comments without AI—There's just so many words—No one has time for that.

Tbf, I'm old enough to have known the world before the world wide web, let alone ChatGPT, and I really don't understand how the hell anyone could write an internally consistent 1000+ page book/series.

Writers are witches.

I've written two books, both well over 100,000 words. While writing the books, I had to read, and re-read them several times, of course. I also spent quite a bit of time researching things, reading parts of other people's books.

The first one I wrote before ChatGPT existed. I finished the second one last year, but I have no idea how to use ChatGPT to write a book, so I didn't try. Besides, writing is fun, why let a stupid machine do it?

chatgpt please summarize all 3 of these

AI in this case meant "Actual Intelligence."

Things were already pretty bad under 'clickbait' slop, AI has certainly exacerbated the issue.

Also most professors try offsetting the cheaters by increasing exam weights, meanwhile people using chatbots finish all their homeworks in like 5 minutes and spend rest of their time studying for the test specifically.

Valuable life skill (being good at bullshitting with little to no preparation)

Those big stacks of flat wood from the before fore?

Heat many boilers to keep turbines spinning with those. The machine gods demand it

The internet is completely different, it's not an LLM.

You people do realize this is likely made-up ragebait, right?

As someone currently in college (soon not to be as I am dumb), I believe it. This indeed is how many do assignments, and also study, thus fully trusting the correctness of the information.

But it's coming from both sides.
Lecturer on programming class told us we can ask questions if we have some. Cool, I did.
So I asked, got "That's a good question, let's find out.", he typed out my question to Gemini, and read back the answer.
This was an introductory class to C, and my question was if you can find how much memory was allocated with malloc. His answer was Gemini's answer which was a concrete no.
Stackoverflow however provided me with better information on the matter: https://stackoverflow.com/a/1281720 (possibly yes)

Also, I've seen him ask and copy paste a number from Gemini. I think it was what a max value is for uint32_t.
Putting 2^32^-1 into a calculator? Nah, let's use an inefficient non-deterministic calculator instead.
Lastly, he added "AI is your most patient teacher".

He did the speaking part of lectures above average, but holy information sources...

You can pick nits that OP's specific quote didn't happen if you want to bother checking. I don't care.

UC San Diego freshmen not prepared for college-level math, writing classes: Report

According to the report, the number of freshmen at UC San Diego who did not meet middle-school proficiency standards in mathematics increased nearly 30 times between 2020 and 2025, despite students having taken the required high school math classes.

The report goes on to say that in 2024, two out of five students who needed to take remedial math classes also had to take remedial writing classes.

Student prompted ChatGPT to write about "homeliness" and not "homelessness."

Made my kids write an essay...the tears OMG the tears!

I can write a solid 10 page well-researched paper out of a library in two weeks. With the internet, I can write that same paper with even better research in one week. With AI tools, I can write that same paper with better readability, conciseness, and overall editing in a couple days.

The difference is that, despite you knowing how to reach the end goal, someone who doesn't understand any of what you do could match your output of "a couple days" in 3-4 days' time. It cheapens the value of actually understanding what you're talking about. Using LLMs might feel more immediately productive, but in the long term it's diluting the skill pool, making it harder to distinguish between deep and shallow knowledge, and allowing (the idea of) merit-based priority to land upon those who don't deserve it more often than it already does (not that I believe our society is largely merit-based anyways).

Sure, the knowledgable can get a lot more done in less time, but the unknowledgable can more easily pretend to be knowledgable, and that technical debt will accumulate heavily over time.