Y'all Need to Read Something Other than Tech Blogs
<rant>
I haven't written a blog post in a while, and I have some interesting things to write about, but today instead of writing about something interesting, I'm going to moralize about something that has been really getting under my skin.
Having spent some time away from reading virtually anything online, I recently started reading blogs again. This is obviously a mistake, and clearly due to some inbuilt failing of my mental constitution, but that is neither here nor there. What I want to talk about today is the terrible state of prose in the tech blog scene, and the inundation of this kind of faux gravitas that I personally associate with LinkedIn. You know the style. Here is a small selection from a few blogs I randomly selected from the past day of top posts on HackerNews.
Once machines start producing code at machine cadence, the model from this book does not break. What changes is the pace: more branches, more commits, more automation, and more surrounding metadata. The traffic gets louder, and the features that keep Git legible under pressure move from "nice to have" to "essential."
The final proof came back with a rather small complaint: the lower counter on the number 8 had closed. Not all eights. Just the 8. JetBrains Mono, sixteen styles, one un-closed path at the foot of the numeral. At screen resolution and at every printer one had checked it on, perfectly fine. At full size on matte paper from Amazon's printers, the counter quietly filled in, and the 8 acquired the silhouette of a 0 with ambition.
Every value in a Rust program has a single owner at any given time. When that owner goes out of scope, the value is dropped — memory freed, deterministically, at a known point in execution. No GC pause, no dangling pointer, no double-free. The compiler enforces all of this statically.
That token had been created for one purpose: to add and remove custom domains via the Railway CLI for our services. We had no idea — and Railway's token-creation flow gave us no warning — that the same token had blanket authority across the entire Railway GraphQL API, including destructive operations like volumeDelete
Now, I'm not necessary saying these were written by generative AI, but they certainly sound like it. It's hard to describe what it is exactly, but at this point if I see a list of negated sentence fragments, I pretty much immediately nope out. No second thoughts, no hesitation, just exit the tab, as it were.
Now, there are sort of two things I want to get at here. The first is, let's assume these are actually humans writing this corporatized pablum. In this case, the solution is easy: pick up a book, a literary magazine, a periodical which actually costs some money and enforces high writing and journalistic standards, whatever. Just read something other than tech blogs! Your writing will be a thousand times better for it, and maybe you'll escape the ouroborus of content generated more and more by AIs, for AIs.
The second is the more likely scenario, which is that I am experiencing a miserable spectrum of varying degrees of AI-driven regression-to-the-mean:
- LLM-requested editing and review, combined with too much trust in its suggestions
- Certain sections iterated on with the assistance of an LLM
- Certain sections entirely generated via prompt, and then tweaked to adjust for style, or to "make them seem more human"
- YOLO ALL PROMPT ALL THE WAY CLAUDE WRITE ME A BLOG POST ON HOW BAD YOUR WRITING IS
If you want to utterly disrespect the people reading your content by not even taking the time to write it, that's your business, but my ask in these cases would be that people at least disclose the level of AI usage in their writing. Just like a translation of a work includes a byline for the translator, if the AI wrote most or a substantial segment of the actual words I'm reading, it should be credited. I'm really burned out on starting to read something that looks like it might be interesting, only to slowly become more and more certain that there's no human in the words.
But, one may argue, it's my content: I drove the prompts, I decided which most-likely token vomit to include in the final product. To this I say, so what? Guess who also was largely writing down someone else's thoughts? That's right, Plato. But you don't see us labeling his books as being written by Socrates, do you? So, even if you feel like you're the Socrates of tech blogs, breaking hitherto untrodden ground with every prompt that you siphon into the machine, you should still have enough respect for your readers to be up front about who wrote them down.
With that in mind, I'd like to propose a set of bylines to put front-and-center on your blog posts to let people like me know whether or not we should even bother to keep reading:
- Author: Me; Editor: Claude Opera 5.6
- Use this one when you wrote most of the content yourself, and then fed it into the normalization machine to get feedback on how you can best sound like the absolute average of everyone else in the entire world
- Author: Me; Co-author: Chat J'ai Pété
- Use this when any section of the text was not originally generated by you. This includes if you attempted to spruce up the LLM's mediocre output
- Author: Geminus; Prompter: Me
- Use this when you didn't write any of it, just used the machine to iterated on its own output or, god forbid, one-shotted something you for some reason expect other people to take the time to read
And also, just to be clear, everything on this blog is written by me, using my own little brain, for better or worse. I'll occasionally ask a friend to look a post over before I publish it, but I neither care about nor seek the input of language models. It's starting to feel like I should add some kind of disclaimer about this: maybe I'll put it in the footer. In a way though, that would feel like capitulation. The default is and should remain that if you put your name on something, the expectation is that you authored it.
</rant>
Another symptom of people reading nothing but tech blogger slop is the weird bubble tech people seem to be living in regarding the opinions of the general populace about having the job-displacement machines shoved down their throats at every possible opportunity. There are certainly exceptions, but essentially everyone I know who isn't in tech is deeply suspicious about generative AI, if not downright hostile. This could of course be due to the fact that I also exist within a bubble defined by where I live, general education level, and so on, but it's remarkably consistent across generations, income levels, and so on. The people I know who are in tech are more split. But it is surprising to see any amount of surprise among the tech crowd to things like:
- Sam Altman's house being attacked twice (or three times?) in 24 hours
- Opinion polls showing that people really dislike AI (e.g. this one)
- Gen-Z dislikes AI more with each passing year
Most people realize on some level the deep harm that social media has done to society, even if they cannot personally extricate themselves from it. With LLMs, we have a new technology being foisted upon us at every possible turn, which, unlike social media, does not even come draped in the trappings of making people's lives better; instead we have ghoulish tech execs relishing in the job cuts, the autonomous drone warfare, the collapse of the commons, and the blood that may yet be spilt on the altar of CAPITAL GAINZ. You'd have to be insane to expect people to be excited about it.