Hey buddy,

Man, Hacker News was pretty wild this Sunday, November 8, 2026. Lots of AI stuff, as usual, but some other interesting bits too. Lemme hit you with the highlights:

AI Wars: Claude vs. Pi & Goldman Sachs Bailing

So, a huge one was this post, "Goodbye Claude, hello Pi". Basically, people are ditching Claude for Pi, saying Pi's GUI is cleaner and hides a lot of the techy stuff in expandable sections. Some folks in the comments were like, "Nah, TUI (text-based user interface) is always worse," but others mentioned Claude Code now has expandable sections too. A common worry was getting banned for using these AI tools in weird ways, especially around API credits.

And get this, right on its heels was the news that Goldman Sachs is actually pulling Claude Code from their bankers and traders! (link here). The main reason floating around in the comments is the huge data leakage risk – you don't want sensitive bank info going to external LLMs. People were talking about needing proper sandboxing for these AI agents, so they aren't running wild on workstations. Though one person pointed out they're only pulling it from *some* employees, not the engineers, suggesting it's more about productivity for certain roles.

Nvidia CEO Jensen Huang's Personal Life Drama

This was a big shocker. Jensen Huang, the Nvidia CEO, wrote a letter to his staff about an affair (read it here). Then there was a follow-up statement from the Nvidia board (here). People were debating whether his personal judgment affects his business judgment. Some comments were like, "Nah, totally separate," while others argued it impacts his credibility or political clout. Someone even joked about his claim of not being able to "walk down the street to get a pizza," saying he's famous for eating street food in Taiwan!

KetchUp: Open-Source SketchUp Alternative in Rust

Okay, switching gears. There's a cool new project called "KetchUp". It's an open-source alternative to SketchUp, built with Rust and WebGPU. So, like, a free 3D modeling tool! A lot of people in the comments were stoked because professional software subscriptions are getting pricey, especially for hobbies. One guy even mentioned he might build his own stuff on top of it, which is pretty neat.

Llama-modchip: Big AI on Small RAM

This one's wild for the AI nerds. Someone figured out how to run a 284 BILLION parameter DeepSeek model on just 64GB of RAM, without a GPU! (check it out). That's insane for running powerful LLMs locally. The comments were buzzing about how cost-effective this could be compared to paying for cloud APIs, with one person saying their DeepSeek bill was only a few bucks a month for agentic coding.

Outlook.com Cracking Down on Email

Heads up if you run your own mail server: Outlook.com is now rejecting emails from domains without DMARC. DMARC is basically an email authentication protocol. It's a pain for small personal domains, but people in the comments agreed it helps the overall mail ecosystem by fighting spam. Microsoft apparently has an authorization system for self-hosted mail servers, which is "stupid but not terribly difficult to set up."

Police Spying Through Ad Tech

This is a bit scary. A Bloomberg piece (link) highlighted how ads in popular apps give police a powerful tool to spy on users. Basically, all that location data and other info collected by ad networks can be accessed by law enforcement. Comments were pushing for network-wide DNS ad-blocking and suggesting tools like Exodus to identify apps with spyware libraries.

Why Some Devs Are Skipping LLMs for Coding

Finally, there was an interesting contrarian take: "Why I will not be using LLMs in my coding". The author talks about the inconsistency of LLMs – one day they're an expert, the next they're an idiot – making them hard to trust. A good point in the comments was treating every LLM session like a "new intern." Another comment made a connection to "Taylorism" in management, suggesting that relying too much on LLMs for code might lead to burnout and a lack of real learning for developers. Food for thought, right?

Anyway, that's the gist of it. Talk soon!