thoughts on calibrating your sources and exercising judgement while consuming online content
there's a lot of noise right now about ai-generated content flooding the internet. most of it isn't wrong. a lot of what's being produced is low-effort, low-density, surface-level content but i keep seeing this framed as a new problem, and i don't think it is.
since i started working a few years ago, one thing i noticed early and continue to experience is how many people apply frameworks they read online or in a book blindly, without situational judgment. they'd drop in a 2x2 matrix, or use some fancy use words, and write/say things that sounded sophisticated but were actually pretty empty. the shallow content problem is old--AI just makes it cheaper to produce.
earlier, creating crap at scale required some effort. you had to sit down, string sentences together, and at least understand what you were writing well enough to write it. now it's a simple prompt to "take this concept and turn it into a blog post." the cost of generating a 1200-word article that says very little has dropped to near zero.
the information density problem got worse, not fundamentally different.
my girlfriend and i were talking about this commonwealth short story competition where the winner was later accused of being AI-written. we debated about whether it was still a worthy piece of literature.
here's my honest take on this. if i read that story and found it compelling without knowing, would i feel cheated afterwards? a bit. but would i think it was a bad story? not necessarily. the judges cut it from whatever 10s of 1000s of entries to one. experienced readers and writers thought it was the best story from its region. that means something.
does that make it okay in a competition? no. competitions have rules for a reason, and submitting AI writing to a "human writing" competition is basically using performance enhancing drugs at the olympics. it's not about whether the output is good, it's about what the competition is actually measuring. if the point is to find the best prompt engineer, fine. if it's to find writers, then no, i don't support it.
a lot of people conflate "ai generated content is often bad" with "ai generated content is inherently bad." these aren't the same.
from a pure utility standpoint, the utility-per-effort ratio of AI varies enormously depending on who's using it and for what. someone who knows what they're doing, using AI to accelerate ops and produce outcomes at 2-3x the norm is fundamentally different from someone copy-pasting the first output they get. the tool doesn't determine the output quality, the judgment of the person using it does.
this is the same as the frameworks problem. the people who applied the 2x2 to everything weren't bad because they used a 2x2. they were bad because they used it without judgment. AI is the same.
there's a separate conversation happening in parallel that gets drowned out by slop discourse. AI is genuinely enabling things that weren't possible before.
the guy who figured out how to treat his dog's cancer by going deep with chatgpt and synthesizing research that no individual vet had the bandwidth to pull together is not slop. that's real leverage for someone who wouldn't have had access to that kind of synthesis otherwise. and it managed to improve the QoL for that poor dog. the other example is thattextflow js implementation someone built with AI assistance, doing layout math that would've required months of specialized work and no-one wanted to do it, but now has applications in the web.
at SpotDraft, we had NLP code from 4-5 years ago that was the best our team could do then. when we threw AI agents at the same problems recently, they wrote code we genuinely couldn't have written five years back. not because we were bad engineers, but because the capability floor has shifted.
at the frontier, people are doing things that weren't possible before. that part of the story exists alongside the slop.
it's not that hard: exercise judgment, same as before, just more actively.
i'd still read AI-generated content if it's good. i wouldn't care whether it was generated or human-written if the information density is high and the thinking is sound. the question i ask of any content is whether the person (or system) that produced it actually understood what they were saying. shallow writing is shallow whether it came from a human or a model.
what i'd stop doing is using "ai generated" as a heuristic for quality in either direction. some of the densest, most useful writing i've read recently was AI-assisted. some of the worst content i've ever seen was written by humans who should've known better.
the web has always required you to calibrate your sources. that calibration is just more important now, and the signal-to-noise problem is harder. find people who exercise real judgment. check whether the thing you're reading actually has depth behind it, or whether it's just words arranged to sound like it does. that was the job before. it's still the job.
narrated by a human (me) and finessed by an AI