Claude 3.7 Sonnet is now available in GitHub Copilot in public preview · GitHub Changelog

Zed: copilot: Add Claude 3.7 Sonnet to Copilot Chat

DeepSeek open source DeepEP – library for MoE (Mixture-of-Experts) training and Inference
Article, Comments

DeepEP is a communication library tailored for Mixture-of-Experts (MoE) and expert parallelism (EP). It provides high-throughput and low-latency all-to-all GPU kernels, which are also as known as MoE dispatch and combine. The library also supports low-precision operations, including FP8.

#cuda #deepseek

seems like some AI agents are getting bullied
https://x.com/skcd42/status/1894375185836306470

this somehow makes me wonder about the future generations

Google Veo 2 has surpassed OpenAI’s Sora and Kling 1.5 Pro as the new leader in Artificial Analysis Video Arena!

Google quietly launched their Veo 2 model via partner services @fal.ai and @freepik (not yet publicly accessible on Vertex). We have observed strengths in rendering people and realistic physics interactions.

Key details regarding Veo 2:

➤ The model is able to generate minutes of 4K video. However, the limited release currently offers 720p video with a 8s maximum duration.
➤ Pricing is $0.50 per second of video generated. This is generally more expensive than other models - a right Google may have earned with its impressive model!
➤ Videos are watermarked with SynthID to ensure it is possible to identify the videos as AI-generated.

Source

Bybit Hack Forensics Report: Bybit

https://docsend.com/view/s/rmdi832mpt8u93s7

⚠️ The forensics investigation of the three signers' hosts suggests the root cause of the attack is malicious code originating from Safe{Wallet}'s infrastructure.

*No indication of compromise was identified within Bybit's infrastructure.

*The investigation is still ongoing to further confirm the findings.

https://x.com/safe/status/1894768522720350673

so here is summary of what happened:
apparently, an employee of a company named Safe{wallet} machine's was compromised (the company named Safe{wallet} announced this in their social account, but didn't specify how it got "compromised" exactly)

my guess is: it was a worm
the attack is attributed to lazarus-group, so my assumption is that, just like their 2017 attack, they made another worm and somehow got it to machine of one of the employees of this company called Safe{wallet}

but the only thing they did: inject a few lines of JavaScript code into the normal code which ByBit was also using, targeting ByBit, not the Safe{wallet}

and it had a very specific trigger condition, which is why literally no one found it out before this happens

thanks to this, ByBit lost around $1.4B, at February 21, 2025; stolen assets were: 401,346 ETH ($1.1 billion) as well as several other iterations of staked ether (stETH)

Bybit CEO Ben Zhou wrote on X that a hacker "took control of the specific ETH cold wallet and transferred all the ETH in the cold wallet to this unidentified address."

SafetyCore: Google Starts Scanning All Your Photos

Google’s awkward moment relates to its SafetyCore, an Android system update that enables on-device image scanning that could do all kinds of things, but is currently focused on blurring or flagging sensitive content. It’s seemingly even more private than Apple’s Enhanced Visual Search, given that it’s all on-device. So we’re told.

you can disable it tho, as it seems
or just switch to GrapheneOS, idk (like a certain someone did in this channel, not me tho)

TypeScript types can run DOOM

https://www.youtube.com/watch?v=0mCsluv5FXA

#Doom (seriously tho, it's hilarious)

and he had to remove all of these limits from the original compiler for this

MakeMake: Generate make files from C source code
Article, Comments

Once MakeMakeMake has built MakeMake you can call it directly.

The MakeMake can also generate the MakeMakeMake script by calling:

./makemake makemakemakemake

SWE-bench for gpt 4.5
(basically this is a benchmark for writing code by the model)

it's not that impressive in fact

LoL it definitely does not worth it in any possible way (in my opinion)

the "creativity" they mention....it's not there??? (or I didn't see it??)

>GPT‑4.5 is a very large and compute-intensive model, making it more expensive than and not a replacement for GPT‑4o. Because of this, we’re evaluating whether to continue serving it in the API long-term as we balance supporting current capabilities with building future models. We look forward to learning more about its strengths, capabilities, and potential applications in real-world settings. If GPT‑4.5 delivers unique value for your use case, your feedback will play an important role in guiding our decision.

ok I take back my words
it does indeed seem like a relatively better models than the previous models