The top 10 results were getting pretty bad even pre-LLM. I remember tons of poorly written articles near the top, just gaming the system with SEO. Everyone was paying some guy on Fiverr $5 an article to churn out hundreds of articles with little regard for accuracy.

We built our own ways to filter these because *we* went to the website and gathered a lot of information on it.

Now, the model slips it in without any verification and casually builds on top of it.

Crucially, by visiting those websites you gathered a bunch of context clues which told you “this source is trash; I shouldn’t trust it and should look somewhere else”. When you ask an LLM, every source is treated as equally valid. You may have gotten an answer from the utmost professional on the subject or an obvious (to you) scam, they are presented the same.

Even if you are in the minority who asks for sources and truly verifies them, they are often made up (don’t exist) or don’t say what the LLM said they said.

Finance LinkedIn is filled with slop posts about "Look at the trading terminal I built. Real math. Here's why it matters." and it's some clearly vibecoded pointless app. Comments are all "Insane project !"

It is funny that they actually added a "Looks like AI Slop" button on the menu for each post.

I'd like to think that the slop-alert button is for their own internal measurements and figuring out much of their userbase is just fake

Seriously, even the "Learn more about limits" link just takes you to the posted page where it literally says nothing about limits or usage. Not even x20 or x10 or whatever, just zero information on what "more usage" actually means. I don't understand why they can't provide how much token usage you get.

Yes they used to show an estimate of how many messages you can send in each tier and for each model.

But past few weeks, I don’t think it was accurate since my usage was going down faster. They nerfed the usage and now aren’t even giving estimates.

Now I'm imagining some kind of timezone-like temperature unit that varies around the world, and a "100 things programmers should know about ___" post about all the intricacies and special cases.

Ooh, and it could shift seasonally so that summer temperatures and winter temperatures can both be in the convenient range. And every locality could shift a different amount and at different times of year according to their local climate.

Next, we can look into reforming those pesky mass and volume units…

I cannot wait for leap degrees.

Didn't expect to see the goalposts moved to "AI can fix bugs but do they even matter"

I'm surprised to hear this. I've found AI incredibly useful for optimizing queries. Give it the EXPLAIN ANALYZE output of your query and it will have no problem optimizing it

Same here, this seems like a setup issue.

If you give the LLM the ability to run queries against prod-like data and review the paths, you have a fully autonomous system that can correctly optimize your ORM/queries. Been doing this for months.

pgAnalyze is a great complementary tool for this and they offer MCP

Immediately left before trying to figure out what is even going on here. Almost every single response is AI generated

I just downloaded and paid for Codex this week because I want to stay on top of the AI tools and understand their capabilities. I've had some good results using 5.6 Sol, although it tends to never want to write any comments (despite modifying the project rules to tell it it MUST), and also occasionally just does a bit of thinking and stops. It'll say "Working through the remaining work" and just ends the chat until I tell it "continue" or something.

This is very anecdotal evidence but I have to rant about it... I tried 5.6 Terra (high) earlier today to fix a bug with a slow page. It just... removed the part of the page that was slow, and made it a client side request (still slow, but not blocking SSR I guess). I tried with Sonnet 5 and it correctly found the issue where an unhandled case was continuing to retry and failing. I am always telling people how the frontier models are SO much more capable than what they may think but this one thing today had me scratching my head at why it would ever do that. It was the first time I experienced the "great, I removed the failing test case" kind of issue.

I got the complete opposite of what you get, sol ultra literally vibed the entire system out for me from one plan mode approval.