Sycophancy in GPT-4o
We have rolled back last week’s GPT‑4o update in ChatGPT so people are now using an earlier version with more balanced behavior. The update we removed was overly flattering or agreeable—often described as sycophantic.
The latest blog posts to take off on Hacker News. 80,845 posts.
We have rolled back last week’s GPT‑4o update in ChatGPT so people are now using an earlier version with more balanced behavior. The update we removed was overly flattering or agreeable—often described as sycophantic.
This report highlights a rarely-discussed but crucially important attack surface: security vendors themselves.
Toyota Motor Corporation (“Toyota”) and Waymo reached a preliminary agreement to explore a collaboration focused on accelerating the development and deployment of autonomous driving technologies.
where my words occasionally escape /dev/null
This Gradient Updates issue explores how much energy ChatGPT uses per query, revealing it's 10x less than common estimates.
The numbers clearly show this is a pointless distraction for the climate movement
Rent's due today, thank heavens I just got paid. I don't set up standing orders, got burned once with unarranged overdraft fees, that wasn't a good mont...
The open-source LLM combines the sequence-modeling skill of a transformer with the inferencing speed of an SSM. IBM Granite will soon adopt key Bamba features.
Here’s a look at what we announced at LlamaCon and how you can get started with our newest releases.
In Which I Try to Maintain Human Supremacy for a Bit Longer
What happens when 4,500 people ask for the same feature? At Firefox, we build it. Tab groups have long been the most requested idea on Mozilla Connect –
AWS’s Sergey Melnik, along with HN commenters matashii and Ants Aasma, have identified the cause of Long Fork in PostgreSQL clusters. As described a 2013 mailing list post , the order in which a PostgreSQL primary makes transactions visible is determined by an in-memory lock. Secondaries, however, make transactions…
Oligo Security reveals AirBorne, a new set of vulnerabilities in Apple’s AirPlay protocol and SDK. Learn how zero-click RCEs, ACL bypasses, and wormable exploits could endanger Apple and IoT devices worldwide — and how to protect yourself.
I’m not talking about skill, knowledge, or convincing a world focused on radical acceleration that optimization is necessary. Performance optimization is hard because it’s fundamentally a brute-force task, and there’s nothing you can do about it. This post is a bit of a rant on my frustrations with code optimization…
I dig into the research on intrinsic motivation. Super interesting subject. Long but hopefully quite thorough.
A couple of years ago, I participated in a study of the HealthFerm Citizen Science group in …
Paying for Kagi today feels a *lot* like paying for HBO back in the cable TV heyday.
Due to a simple bug, Screen Studio app did generate over 2 petabytes of network traffic
I've studied the effects of electromagnetic fields on patients with carotid stent implants during my PhD. I've never talked with any patient nor seen any such medical devices even to this day, fifteen years later.
It would give the powerful a dangerous new route to manipulate platforms into removing lawful speech that they simply don't like, with President Trump himself saying that he would use the law to censor his critics. The bill passed the Senate in February, and it now heads to the president's desk.
In numerous applications of machine learning, Hessians and Jacobians exhibit sparsity, a property that can be leveraged to vastly accelerate their computation. While the usage of automatic differentiation in machine learning is ubiquitous, automatic sparse differentiation (ASD) remains largely unknown. This post…
It's waiting for Godot and eventually gives up.
Google pays for 80%+ of the development budget of all major web browsers, and soon, the US will force Google to stop supporting their…
The majority of the traffic on the web is from bots. For the most part, these bots are used to discover new content. These are RSS Feed readers, search engines crawling your content, or nowadays AI bo
QWEN CHAT GitHub Hugging Face ModelScope Kaggle DEMO DISCORD Introduction Today, we are excited to announce the release of Qwen3, the latest addition to the Qwen family of large language models. Our flagship model, Qwen3-235B-A22B, achieves competitive results in benchmark evaluations of coding, math, general…
Why we moved Architect of Ruin from Rust and Bevy to C# and Unity in January 2025, and how the three-day experiment that decided it went.
HN Discussion
11 months in The Side Hustle From Hell