The same can be said about Apple. Several companies have complained about them taking a meeting with apple, presenting their product, only to have Apple then rip it off and build it in house. To say nothing of sherlocking.
I'm not calling it slop because it's long. I'm calling it slop because it's poorly written and reeks of AI with no editing.
Instead of
> Is the loop just more attempts? One objection deserves an answer up front. A verification loop spends extra inference per task, so is the lift just a bigger compute budget in disguise? Partly it has to be. The loop does more work. But the retries the benchmark grants are blind: the model sees a failure signal and guesses again. The loop’s iterations are guided by evidence from the running application, which is a different kind of attempt, not just another one. Whether guided iteration beats an equal budget of blind retries at matched cost is exactly the ablation this framing demands, and it is planned for a future post: DeepSeek alone with a larger retry budget, against DeepSeek with the loop, dollar for dollar. Until that runs, read the results below with this open question in mind.
It could have been
> These loops are not just retries. Each iteration provides the model with evidence from the previous run. Some of the uplift may come from the extra tokens, so a follow-up post will compare the guided loop against cost-matched blind retries.
We can argue over exact wording, but the original is far too long.
Or the point about "measuring cost honestly". It's not clear why you wouldn't be using the published rates and do the basic multiplication yourself. There's nothing subtle about this, and it doesn't need to a whole paragraph.
Or the people writing this could spend more effort to make it not slop. If they can’t be bothered, I won’t waste my time figuring out if this is worth it, there’s are 1000 other articles to read and techniques to try.
And if this is worth a look, I’m sure I’ll hear about it again from someone who wrote it better.
Well I don’t know what the method is, there’s a mention of ironbee, but I’m not willing to spend the time digging through this to figure out what’s needed and how to set it up.
Nobody is forcing you to play a game with kernel anti-cheat.
I want to play the game, not deal with mass cheating. Unless the people against kernel-level anti cheat can provide an alternative with similar level of protection, there isn’t anything to discuss as far as I’m concerned.
One thing that appears to be missing is any mention of non-heterosexual couples, some of which are biologically incapable of having their own children and it's unclear how adoption or surrogacy gets counted in here.
And I think it's fair to say that in the US non-heterosexual people are overwhelmingly on the left, for fairly obvious reasons.
Why would I be missing out on learning? If I want to improve my writing, why does it matter that if it’s Word, ChatGPT, or a teaching telling me that my spelling is wrong? Or that I keep making comma splice errors? Or that I have run on sentences?
> You need to hand it in so that the teacher understand where to put in more effort, but if it wasn't for that need, they'd probably have you throw it away after writing it.
I am not sure how your literature classes went, but all of my essays were graded and feedback was provided to me specifically so I can get better. Perhaps my previously reply was too long winded, but feedback on your essay on how to improve it is the exact use case I gave as an example.
We'll put aside where AI usage is a skill that needs to be taught (which I think there is definitely room for teaching people how to use it effectively as a tool) since that isn't what the discussion is about.
The article's talking about the use of AI learning rather than learning how to use AI.
> If AI does that work for them, then they simply don't learn.
I agree, and I think the original commenter would agree too given that this doesn't sounds like moderation.
The no-ai end is "you write the whole essay yourself" the all-in end is "you give the ai and idea and have it write the essay". The moderation approach is somewhere in-between and it could very well be "you write the essay and ask the AI to proofread and coach you through it".
> It's largely the same reason we don't let children use calculators when they're learning basic arithmetic. Calculators exist, and are useful, but they're awful in a teaching environment.
Yes, having the ai act as a calculator when you need to learn and prove you can do it is a bad use of it. Having the Ai double check your work to catch errors, point out when you make the same mistake over and over, or ask it to walk you through another example are all productive uses.