this is a great example of how obviously silly psychology research is a lot of the time should you treat unwanted thoughts as irrelevant or dig in to them? depends highly on context a similar analogous question is: “what’s the next token in a sequence?” it’d be obviously
RESEARCH
-

New AI model matches 4o and Claude in writing without refusals
By
–


Nothing groundbreaking on capabilities so far, it's more that it's at least in the same league as 4o or Claude at writing without subject-related refusals. An Onion-style Buzzfeed listicle about weapons you can make in a toilet stall may not be funny, but it's not a crime:
-
Intelligence, Experience, and the Path to AI Understanding
By
–
Hypothetically yes given an infinite amount of time and the right propensities / life experiences. But the hypothetical there is doing a lot of work…the path to relativity is very complicated and not well captured by IQ graphs or benchmarks
-
Can Current AI Systems Achieve Revolutionary Scientific Breakthroughs?
By
–
I think that’s very unlikely. A good parallel thought experiment is think of a smart high schooler, maybe the smartest high schooler you’ve ever met who’s taken basic high school physics. Would they come up with relativity?
-
Can AI Discover Relativity From Early 1900s Data Alone?
By
–
I proposed a version of this at a dinner to some high level AI (train it only on data up to the early 1900s and see if it comes up with relativity) They said there wasn’t enough data. Not sure if that applies to high school too but it might
-
Fine-tuning’s Surprising Power: Knowledge Retention Through Style Change
By
–
Wow it really has been that long 😐
The big thing I didn’t realize is that an assistant was just a finetune away. That is the surprising thing I was really missing. I still find it surprising today, that you can just change the style so dramatically but retain the knowledge. -
Three Essential LLM Books Covering Different Perspectives
By
–
Spotted this on Amazon, and it was such a nice coincidence and spot-on recommendation that I had to share it. As someone who wrote one and read the other two, what I like about this selection is how they cover LLMs on very different levels, with basically no overlap (besides them
-
TorchInductor Dynamic Kernel Generation Advances Tensor Compilation
By
–
TorchInductor supports dynamic kernel generation, and is likely the state of the art at symbolic Tensor type systems. long gone are the days of static shape tensor compilers
-

2025 LLM Roadmap: From Fundamentals to Production Applications
By
–
Here’s the 2025 LLM roadmap 1. Code and train your own LLM to really understand the fundamentals
2. Train models more conveniently using production-ready libraries
3. Learn about the big-picture considerations for real-world LLM/AI apps -

Embedding Fine-Tuning Guide for RAG Systems
By
–
RAG Embedding Fine-Tuning Guide A practical guide demonstrating embedding model fine-tuning for RAG systems, featuring LangSmith monitoring and RAGAS metrics for performance evaluation. Check out the complete implementation https://
github.com/apatti/AIEBoot
camp/blob/main/09_Finetuning_Embeddings/Fine_tuning_Embedding_Models_for_RAG_using_RAGAS.ipynb
…
