@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

Just random ramblings of a bored contemplating employee. I'm with the crowd who don't think AI will be replacing the workforce soon. Nope. But the workers who do know how to use AI to their best interests will definitely have the upper hand. Me included. So in my current job, it's a WFH setup and doesn't require much invasive privacy trackers at all. So I'm free to use AI to *"do x for me"* basically. It runs code, walks me through tools/software I have to familiarize myself with. But I woul

**@sama (Sam Altman) posted the following key content related to AI in the past 24 hours (late March 3, 2026 GMT):** - **Main thread post (ID: 2028640354912923739)**: Reposting an internal update about

@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

**@ylecun (Yann LeCun) did not post any content directly discussing AI, LLMs, or machine learning in the past 24 hours (since March 3, 2026).** **His on

**@ylecun (Yann LeCun) did not post any content directly discussing AI, LLMs, or machine learning in the past 24 hours (since March 3, 2026).** **His on

@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

**@sama (Sam Altman) posted the following key content related to AI in the past 24 hours (late March 3, 2026 GMT):** - **Main thread post (ID: 2028640354912923739)**: Reposting an internal update about

@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

**@ylecun (Yann LeCun) did not post any content directly discussing AI, LLMs, or machine learning in the past 24 hours (since March 3, 2026).** **His on

@GaryMarcus posted the following key posts related to AI, LLMs, or machine learning in the past 24 hours (as of March 4, 2026): 1. **New study on sycophantic AI in LLMs** (March 3, 16:35 GMT): "New study that everyone who uses LLMs should read.

The release of qwen just makes me shake my head in disbelief. I can get coding help by asking natural language questions like I would to a real human - without even needing internet. It’s fucking insane.

First things first, claude is much better at just talking, it understands context, has jokes and tries to swing you the right way if you’re spiraling or wasting time instead instead of just fueling it like chatGPT does, it’s genuinely more fun to talk to. However all the fun ends here. As soon as you need it to be actually helpful, its starting to get annoying fast, it hallucinates a lot, you have to specifically ask it to use tools in the prompt, otherwise it just makes stuff up, or tried to

The OS can be used here: [WebOS 1.0](https://qwen4bwebos.tiiny.site/) Prompt used was "Hello Please can you Create an os in a web page? The OS must have: 2 games 1 text editor 1 audio player a file browser wallpaper that can be changed and one special feature you decide. Please also double check to see if everything works as it should." Prompt idea thanks to /u/[Warm-Attempt7773](https://www.reddit.com/user/Warm-Attempt7773/) All I did was to ask it to add the piano keyboard. It e

https://preview.redd.it/4fjzkqelxumg1.png?width=1178&format=png&auto=webp&s=c6b0015cec7f0970b412b41d52548a90e949c13b Thank him for his contributions to local LLM

https://x.com/openai/status/2028893702865989707?s=46

Hey folks! Long-time lurker, first time poster. I’m a PhD student, and I’ve been wondering: how much time do you actually spend just trying to reproduce ML papers? Even when the code is available, it can take days (or weeks!) to get everything running—tracking down missing hyperparameters, figuring out weird environment issues, or just dealing with stuff that’s buried in an appendix. So I’m genuinely curious: \+ How much time do you lose each week just getting baselines or prior work