From what I understood, the gist of it is that uses a bunch of skills in a loop to help LLMs solve the tasks better via browser-testing, debugging, visual-testing, etc..
It is marketing for some SaaS thing. However, I think something like this, "verification loops with some sort of tricks", is what Anthropic and "OpenAI" do behind the scene with their interfaces.
Fair on the length. The verification loop took DeepSeek from ~20 to ~80 on Web-Bench, matching Opus at ~7x less cost. Curious if the method holds up for you.
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.
Zero engagement with anything the post actually says. You've spent more words performing your distaste than it would've taken to read the thing. Recycling slop as a catch all for "I didn't read it" is the real content free contribution here.
LOL, you could at least write your comments yourself.
I said I didn't read (so much for the big gitcha) because if its not designed for human consumption that means it's not worth human time.
I bet someone actually respecting their audience will present findings in a way that's.... palatable, for the lack of the better word.
(I spent some time commenting to add a voice of disagreement for enshittification of the publishing part of the internet. I'm entitled to it as much as you feel entitled to posting your comments)
It's not too long for humans, but it seems that it's too long for you maybe because you don't know how to read :) You didn't finish it, so you assumed nobody could. Don't speak for all humans based on your own reading.
Haters like you, who skim other people's comments and pass them off as their own thoughts, are everywhere. I don't have time for it. Try you chance with other posts.
It is not the AI slop, it is the benchmark prepared and done by humans which people like you can't understand. You still keep posting slop comments without ZERO concrete argument.
And also, I'm not saying you said you skimmed my comments. It doesn't need to be said, it is obvious that you wrote those comments without opening the link by just reading other's comments. That's it!
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.
Expressing some frustration with the authors lack of consideration when announcing something is fairly natural, tbh.
Posts like this break the social contract, and honestly show a lack of both care and consideration for communicating the subject matter AND a flagrant disrespect for anyone who would consume this.
I’m sure there are some good ideas here, but it’s hard to sift through the lack of applied ai and lack of communication skills to get to them
@neetle I get the frustration, but reaching for "slop" as the default reaction to anything you find too long isn't all that natural. Happy to hear specific feedback on what didn't land, that's more useful to me than the label.
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.
Ok, even though "poorly written" is just your bias to me, at least you could provide some concrete critiques. It seems that you listened to the others, had the AI read the blog post, and asked it to write something concrete based on that.
> 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.
Because, while Opus provides cost in the API response, DeepSeek doesn't, but just token usage. So, cost is calculated based on used tokens for DeepSeek. We have added this section, because draft version of the post got feedback on this.
Nice try with the metaphor, but it is just useless. I'm pretty sure that you didn't event open the link, but just read the other comments, saw the word slop and wrote something to get attention. There are people like you everywhere. If you had actually read the post, you would point to a real problem in it. You can't, because you didn't read it.
I don't control your thinking nor do I care, you're free to project your fobias and traumas on everyone else.
Why do you expect me to spend even more time to laboriously show you examples of the slop? The other comment already did, I only engage with the actual work, I wasted too much time on it already.
Better luck next time. Oh, and a free advice, when someone rejects something you posted, a thing, immaterial, and you recoil with personal attacks at anyone who didn't like it, this is one of the ways you show it's not worth investing more effort into discussing with you or your agents.
> I’m not willing to spend the time digging through this to figure out what’s needed
That's the neat part. You don't... need to do this yourself nowadays. AI agent is perfectly capable to figure out any new open source thing that pops up daily now. You can just give it the link to the repo and say "Set this up for me. I wanna play with it."
From what I understood, the gist of it is that uses a bunch of skills in a loop to help LLMs solve the tasks better via browser-testing, debugging, visual-testing, etc..
The skills can be found here:
https://github.com/ironbee-ai/ironbee-devtools-skills/tree/m...
Question that was left unanswered:
How much longer does a 4x DeepSkeep loop took compared to Opus?
Seems interesting, but buried under a mountain of written slop.
It is marketing for some SaaS thing. However, I think something like this, "verification loops with some sort of tricks", is what Anthropic and "OpenAI" do behind the scene with their interfaces.
Fair on the length. The verification loop took DeepSeek from ~20 to ~80 on Web-Bench, matching Opus at ~7x less cost. Curious if the method holds up for you.
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.
You could literally send it to an ai
Absolutely not.
If someone vomits the slop that must be fed to a robot to make sense of it, even if there's something of value.
It's just so rude, like a professor entering the class, emptying a carton box full of papers on the floor, flipping a bird and leaving without a word.
It's possible there's something good in it.
Zero engagement with anything the post actually says. You've spent more words performing your distaste than it would've taken to read the thing. Recycling slop as a catch all for "I didn't read it" is the real content free contribution here.
LOL, you could at least write your comments yourself.
I said I didn't read (so much for the big gitcha) because if its not designed for human consumption that means it's not worth human time.
I bet someone actually respecting their audience will present findings in a way that's.... palatable, for the lack of the better word.
(I spent some time commenting to add a voice of disagreement for enshittification of the publishing part of the internet. I'm entitled to it as much as you feel entitled to posting your comments)
It's not too long for humans, but it seems that it's too long for you maybe because you don't know how to read :) You didn't finish it, so you assumed nobody could. Don't speak for all humans based on your own reading.
Haters like you, who skim other people's comments and pass them off as their own thoughts, are everywhere. I don't have time for it. Try you chance with other posts.
I read people's work, I stop reading AI slop that people say it's their work.
I didn't say I skimmed your comments, no need to lie if you don't have a good response.
You're too enraged people expect some baseline quality from whatever you post. Why is that so much?
It is not the AI slop, it is the benchmark prepared and done by humans which people like you can't understand. You still keep posting slop comments without ZERO concrete argument.
And also, I'm not saying you said you skimmed my comments. It doesn't need to be said, it is obvious that you wrote those comments without opening the link by just reading other's comments. That's it!
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.
Or just don't read the blog post and stop posting your slop here so everyone will be happy
Expressing some frustration with the authors lack of consideration when announcing something is fairly natural, tbh.
Posts like this break the social contract, and honestly show a lack of both care and consideration for communicating the subject matter AND a flagrant disrespect for anyone who would consume this.
I’m sure there are some good ideas here, but it’s hard to sift through the lack of applied ai and lack of communication skills to get to them
[dead]
If you can't take a constructive critique of something you posted..... maybe don't post so you'll remain happy?
I hope you can take your own advice?
Take care!
@neetle I get the frustration, but reaching for "slop" as the default reaction to anything you find too long isn't all that natural. Happy to hear specific feedback on what didn't land, that's more useful to me than the label.
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.
Ok, even though "poorly written" is just your bias to me, at least you could provide some concrete critiques. It seems that you listened to the others, had the AI read the blog post, and asked it to write something concrete based on that.
> 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.
Because, while Opus provides cost in the API response, DeepSeek doesn't, but just token usage. So, cost is calculated based on used tokens for DeepSeek. We have added this section, because draft version of the post got feedback on this.
@subscribed To make a critique, first you must READ. Otherwise, your comments will be slop
Its enough to skim far enough to see it's just slop.
I don't need to eat entire bowl of the burnt soup with flys floating in it to know its a burnt soup with insects in it.
Nice try with the metaphor, but it is just useless. I'm pretty sure that you didn't event open the link, but just read the other comments, saw the word slop and wrote something to get attention. There are people like you everywhere. If you had actually read the post, you would point to a real problem in it. You can't, because you didn't read it.
I don't control your thinking nor do I care, you're free to project your fobias and traumas on everyone else.
Why do you expect me to spend even more time to laboriously show you examples of the slop? The other comment already did, I only engage with the actual work, I wasted too much time on it already.
Better luck next time. Oh, and a free advice, when someone rejects something you posted, a thing, immaterial, and you recoil with personal attacks at anyone who didn't like it, this is one of the ways you show it's not worth investing more effort into discussing with you or your agents.
What I'm expecting is stop spamming here with your slops, that's all.
> I’m not willing to spend the time digging through this to figure out what’s needed
That's the neat part. You don't... need to do this yourself nowadays. AI agent is perfectly capable to figure out any new open source thing that pops up daily now. You can just give it the link to the repo and say "Set this up for me. I wanna play with it."
until you realize people are putting malicous instructions in the projects now, at least a few anti-Ai coders see this as an act of defiance
I don't think that kind of people builds AI tools. Also you can and should sandbox your AI when trying out something sus.