Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

GPT-5

openai.com

Post reply on HN
Print view View original

581–590 of 1001 posts

Re: GPT-5

#581

Post by Traubenfuchs »

andai wrote: post #315

So models are getting pretty good at oneshotting many small project ideas I've had. What's a good place to host stuff like that? Like a modern equivalent of Heroku? I used to use a VPS for everything but I'm looking for a managed solution. I heard replit is good here with full vertical integration, but I haven't tried it in years.

Set up a free kubernetes cluster on the always free tier of oracle cloud with terraform.

4 nodes with 1 cpu and 6 GB RAM each: that's PLENTY for small project ideas. You also get plenty of free storage/DB options.

After having learned to do this once, creating and deploying a new app under your subdomain of choice should take you no more than a few minutes.

Re: GPT-5

#584

Post by syntaxbush »

wrcwill wrote: post #567

ugh still fails my test prompt: https://chatgpt.com/share/689507c7-5394-8009-b836-c6281a246e... "Assume the earth was just an ocean and you could travel by boat to any location. Your goal is to always stay in the sunlight, perpetually. Find the best strategy to keep your max speed as low as possible" o3 pro gets it right though..

Mine "thought" for 8 minutes and its conclusion was:

>So the “best possible” plan is: sit still all summer near a pole, slow-roll around the pole through equinox, then sprint westward across the low latitudes toward the other pole — with a peak westward speed up to ~1670 km/h.

Is this to your liking?

Re: GPT-5

#585

Post by seanw265 »

levocardia wrote: post #366
surround wrote: post #297

GPT-5 knowledge cutoff: Sep 30, 2024 (10 months before release). Compare that to Gemini 2.5 Pro knowledge cutoff: Jan 2025 (3 months before release) Claude Opus 4.1: knowledge cutoff: Mar 2025 (4 months before release) https://platform.openai.com/docs/models/compare https://deepmind.google/models/gemini/pro/ https://docs.anthropic.com/en/docs/about-claude/models/overv...

with web search, is knowledge cutoff really relevant anymore? Or is this more of a comment on how long it took them to do post-training?

Knowledge cutoff isn’t a big deal for current events. Anything truly recent will have to be fed into the context anyway.

Where it does matter is for code generation. It’s error-prone and inefficient to try teaching a model how to use a new framework version via context alone, especially if the model was trained on an older API surface.

Re: GPT-5

#586

Post by KoolKat23 »

highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

In my opinion, it'll mirror the human world, there is place for multiple different intelligent models. Each with their own slightly different strengths/personalities. I mean there are plenty of humans that can do the same task but at the upper tier, multiple smart humans working together are needed to solve problems as they bring something different to the table. I don't see why this won't be the case with super intelligence at the cutting edge. A little bit of randomness and slightly different point of view makes a difference. The exact same two models doesn't help as one would already have thought of what the other was thinking already

Re: GPT-5

#587

Post by jonathan_h »

arcumaereum wrote: post #143

In terms of raw prose quality, I'm not convinced GPT-5 sounds "less like AI" or "more like a friend". Just count the number of em-dashes. It's become something of a LLM shibboleth.

I am a big fan of using the em-dash.

I won't argue that I always use it in a stylistically appropriate fashion, but I may have to move away from it. I am NOT beating the actually-an-AI allegations.

Re: GPT-5

#588

Post by charlie0 »

Not so sure about the behind the scenes "automatic router". What's to stop OpenAI from slowing gimping GPT-5 over time or during times of high demand? It seems ripe for delivering inconsistent results while not changing the price.

Re: GPT-5

#589

Post by kgeist »

jumploops wrote: post #358

Is GPT-5 using a new pretrained base, or is it the same as GPT-4.1? Given the low cost of GPT-5, compared to the prices we saw with GPT-4.5, my hunch is that this new model is actually just a bunch of RL on top of their existing models + automatic switching between reasoning/non-reasoning.

GPT-5's knowledge cutoff is September 2024 so my first thought was they used GPT-4's pretrained base from 2024 and post-trained it additionally to squeeze those additional +5% on the benchmarks. And added the router.

Re: GPT-5

#590

Post by cchance »

highfrequency wrote: post #331

It is frequently suggested that once one of the AI companies reaches an AGI threshold, they will take off ahead of the rest. It's interesting to note that at least so far, the trend has been the opposite: as time goes on and the models get better, the performance of the different company's gets clustered closer together. Right now GPT-5, Claude Opus, Grok 4, Gemini 2.5 Pro all seem quite good across the board (ie the…

They have to actually reach that threshold, right now their nudging forward catching up to one another, and based on the jumps we've seen the only one actually making huge jumps sadly is Grok, which i'm pretty sure is because they have 0 safety concerns and just run full tilt lol
Post reply on HN