🔥The 9th Round of Easy Loan, Earn $40 Reward is in progress❗️ ⏰ Promotion Period: January 15th - Feburary 15th, 2025 👉 Register now and check more details at gate.io/campaigns/358

Some nice pushback on my GPT-2030 post by xuan (ɕɥɛn / sh-yen), with lots of links!

This is the way to adapt LLM context to different inputs

PSA: If you don't want your X posts used to train Grok, you now have to explicitly opt-out. Go to x.com/settings/grok_… and uncheck the box. If link doesn't work, go to Settings->Privacy and Safety->Grok

What I'm interested is the nature of this weirdness. How can it be true that while it can answer questions that most college graduates can't, it can't answer questions that most children can? How can you not think this is interesting? Why would you want to ignore this question?

Unsubscribed

Excellent read on the challenges of using Kubernetes for developer environments. Gitpod do great work.

this really fits my feeling about "reasoning models"… they are more like super-human "guesser" or "fact-retriever & sequencer", they can magically recall obscure facts, and then try to combine them properly.

Microsoft’s strategy of not liberally licensing their python language server (pylance) seems prescient given its usefulness for ai coding agents. Predication: ai coding agent companies will build or buy language analysis tools and augment them further than just type checking.

The edge hyperscalers still have is that reliability is feature #1

This seems to be the approach coding agents are converging on, see also Claude Code and its thinking tool. The ensemble step is unique though.

Great work documenting how chatbot arena is gamed by the big LLM providers