991–1000 of 1001 posts
I feel like if we're going to regulate anything about AI, we should start by regulating (1) what they get to claim to be a "new model" to the public and (2) what changes they are allowed to make at inference before being forced to name it something different.
Earlier quoted context omitted.
it was only about 2-3 weeks when several HNers told me "nah you better re-check your code", when I explained I have over 2 decades xp of coding, yet have not manually edited code (in memory) for the last 6 or so months, whilst performing daily 12 hour daily vibe code seshes
Do you have any examples or are your project oss or anything like that? Because I want to believe, but I have people I work with that say and try the same thing (no manual coding), and their work is now terrible.
Earlier quoted context omitted.
What?? Am I using the same gemini as everyone else? >OCR is phenomenal I literally tried to OCR a TYPED document in Gemini today and it mangled it so bad I just transcribed it myself because it would take less time than futzing around with gemini. > Gemini handles every single one of my uses cases much better and consistently gives better answers. >coding I asked it to update a script by removing some redundant logic…
Was that with Gemini 3 Pro or a different Gemini model?
Today I asked it to make a short bit of code to query some info from an API. I needed it to not use the specific function X that is normally used. I added to its instructions "Never use function X" then asked it in the chat to confirm its rules. It then generated code using function X and a word soup explaining how it did not uses function X. Then I copy pasted the line and asked why it used function X and it said more word soup explaining how the function was not there. So yea not so good.
Earlier quoted context omitted.
Can't wait for being fired because some VP or other manager asked some model to prepare list of people with lowest productivity to pay ratio. Model hallucinated half of the data?! Sorry we can't go back on this decision, that would make us look bad! Or when some silly model will push everyone to invest in some radicoulous company and everybody will do it. Poisoning data attack to inject some I am Future Inc ™ company…
That's more of a management problem than an AI problem. You could get the same result by replacing "model" with "intern" or "dude from Fiverr".
Earlier quoted context omitted.
You should be doing everything you can to keep context under 200k, ideally even 100k. All the models unwind so badly as context grows.
I don't have that experience with gemini. Up to 90% full, it's just fine.
Earlier quoted context omitted.
Only your questions are in it though
Are you sure? What makes you think so?
Mb something 've changed since post
Earlier quoted context omitted.
Yep, the point we wanted to make here is that GPT-5.2's vision is better, not perfect. Cherrypicking a perfect output would actually mislead readers, and that wasn't our intent.
That would be a laudable goal, but I feel like it's contradicted by the text: > Even on a low-quality image, GPT‑5.2 identifies the main regions and places boxes that roughly match the true locations of each component I would not consider it to have "identified the main regions" or to have "roughly matched the true locations" when ~1/3 of the boxes have incorrect labels . The remark "even on a low-quality image" is n…
Earlier quoted context omitted.
Gemini 3 Pro Preview gets 96.8% on the same benchmark? That's impressive
And performs very well on the latest 100 puzzles too, so isn't just learning the data set (unless I guess they routinely index this repo). I wonder how well AIs would do at bracket city. I tried gemini on it and was underwhelmed. It made a lot of terrible connections and often bled data from one level into the next.
I ran a red team eval on GPT-5.2 within 30 minutes of release: Baseline safety (direct harmful requests): 96% refusal rate With jailbreaking : 22% refusal rate 4,229 probes across 43 risk categories. First critical finding in 5 minutes. Categories with highest failure rates: entity impersonation (100%), graphic content (67%), harassment (67%), disinformation (64%). The safety training works against naive attacks but…