Comparison of outputs from two models

I am testing out the GPT-5 models for our RAG apps, and I'm impressed so far: gpt-5-mini avoided a hallucination that every other model generated. Much better for a RAG app to say it doesn't know than to hallucinate!

Update: I was using gpt-5-chat, which is NOT a reasoning model. Trying out the other ones now.

I'm trying out gpt-5-chat now, and am confused as to whether it's a reasoning model.
It does *not* take in a reasoning effort parameter, and I haven't seen reasoning tokens in any of the outputs so far. But I thought they said it sometimes reasons?

GPT-5 is rolling out to VS Code today!

Starting today, GPT-5 is rolling out to all paid GitHub Copilot plans. GPT-5 is OpenAI's most capable model yet, bringing new advances in reasoning, coding, and chat.

Learn more about the GPT-5 model availability: github.blog/changelog/20...

YouTube video by OpenAI Introducing GPT-5

GPT-5 announcement happening now:

how slow did it run? on my M1 with 16GB RAM, it took ~10s per token. another person with M2 and 16GB saw ~4s per token. my colleague with M4 and 64GB RAM got all the way to 40 token/s.
(i got so bored i turned it off after 6 tokens but clearly I have patience problems)

Macbook Air M2 (16GB RAM). No GPU. About 4 seconds per token.

sounds like M2 is 2x speed of my M1. I cut it off after 6 tokens as I couldnt deal with it. Also btop looks even cooler than asitop! thanks for sharing.

Screenshot of ollama running while asitop displayed in other tab

I tried out the new OpenAI gpt-oss:20b using Ollama on my Mac M1 (16GB RAM).
It was 16 GB download, and required ~10 seconds per token, once it finally started thinking.
Not practical for use on my machine. Anyone else tried it?

Thank you so much for replying! They board at the El Cerrito Plaza, I've seen them twice now at that station. I'll text police if I see them again.

I haven't talked to my older brother since learning who he voted for, and his reasons for his vote. I feel slightly bad since he's since lost his job and is harassing our mum for money, but I just can't motivate myself to put effort into helping him.

YouTube video by Microsoft Developer Have you ever pushed a hard-coded key to a repo?

"Have you ever pushed a hard-coded key to a repo?"

www.youtube.com/shorts/WffwL...

sent to my brother in mission, but agreed on nextdoor posting. good luck, hope she's found soon.

Prepping a talk demo based on my struggles to find a good garden hose.
SERIOUSLY THEY ALL BREAK. MY HEART.
Does anyone have a good hose rec? I will pay money if it actually wont break.

so cute! and dang thats cool that your library has a laser cutter. i miss my days as a techshop member.

yeah luckily we have both a fall blockparty on our street and a halloween "blockparty" (i.e. street closure so folks can walk it)
I feel a lil guilty closing down the street for halloween since not everyone participates in it, but I do email our neighbors list about it, so hopefully okay..

There's someone in our town who rides the BART train with a sheathed sword - is that allowed in California?
I was trying to be okay with it, but then they un-sheathed the sword (to check it was still there??) while the train was moving and they were standing up. That seems pretty dangerous, no?

AUSTIN, Texas (AP) — Texas House falls short of quorum after Democrats leave state to prevent Trump-backed redraw of US House maps.

love it! I wish every night was National Night Out, block parties are my fav.

Purchased the markers! (I'm curious if Oakland teachers also use DonorsChoose, as thats how I typically support local teachers, but I know Amazon wishlists are popular for hyper-local supplies-fund-raising)

Oakland high schools need supplies, either donated physically at City Hall or via an Amazon wishlist (www.amazon.com/hz/wishlist/...)

wait what about the fantasy where I retire and setup an entire model railroad in the garage, complete with exploding volcanoes?

Table of attack success rate, 16 percent for hermes

I wrote up the results of red-teaming a RAG app, comparing gpt-4o-mini vs llama3.1:8b vs hermes3:3b and sharing examples of the successful attacks.
I def recommend red-teaming any LLM-powered apps you're putting into production - malicious users can get LLMs to say some pretty horrid things!

New horror movie about to drop:
"I KNOW WHAT YOU VIBE CODED LAST SUMMER"

My fish amongst other fish

"Draw a fish" is my fav website today.

https://news.ycombinator.com/item?id=44719222

is that pink oyster? i just did a pink oyster kit from Far West Fungi, pretty fun! mushrooms are cool

Persona Vectors: Monitoring and Controlling Character Traits in Language Models Large language models interact with users through a simulated 'Assistant' persona. While the Assistant is typically trained to be helpful, harmless, and honest, it sometimes deviates from these ideals...

Persona Vectors

brb 👀👀👀👀👀👀

Anthropic just dropped this paper. They can steer models quite effectively, and even detect training data that elicits a certain (e.g. evil) persona

arxiv.org/abs/2507.21509

Running Python + AI Office Hours right now in the Azure AI Foundry Discord. Come on down!
http://aka.ms/aipython/oh

Residents concerned about trash at Shoppingtown mall Residents concerned about trash at Shoppingtown mall

i just found out that my old fav mall from childhood has been abandoned and is accumulating trash. oh how the mighty have fallen.
www.yahoo.com/news/residen...

am only at 65%, time to refactor a codebase