In 2020, OpenAI showed that language models are few-shot learners. In 2022, ChatGPT was released. By 2025, Google DeepMind demonstrated that video models are zero-shot learners and reasoners. What’s next?
LLMS
-
Quarantined LLM Systems Modeled as Tool Calls
By
–
Yeah, the quarantined LLM stuff can absolutely be modeled as tool calls instead
-
LLM Researchers’ Knowledge Gap About Language Fundamentals
By
–
Ironically, LLM researchers know nothing about language.
-
Generative AI’s Widespread Structural Limitations Essay
By
–
of course they are (structured) representations. my recent relevant essay: https://
open.substack.com/pub/garymarcus
/p/generative-ais-crippling-and-widespread?r=8tdk6&utm_campaign=post&utm_medium=web&showWelcomeOnShare=false
… -
Tech Industry Overpromises: LLM Hallucinations Still Unsolved
By
–
How the tech industry rolls: Almost a million people viewed this May 2023 tweet from Mustafa Suleyman promising that “LLM hallucinations will be largely eliminated by 2025”; many of people believed it (and believed many similar promises from the tech industry). I didn’t — I
-
Video Model Limitations and LLM Improvement Potential
By
–
In both cases you can come up with edge-cases that expose their limitations – with LLMs may or those limitations see overcome as the models improve, will be interesting to see if that happens for video models
-
Chain of Thought vs True Reasoning in LLM Next Token Prediction
By
–
Feels about equivalent to chain of thought in next token guessing LLMs being used for "reasoning" to me
-
100+ Free Open Source AI Agents and RAG Tutorials
By
–
Stay tuned for more such interesting posts → @Saboo_Shubham_ I have created 100+ AI Agents and RAG tutorials, 100% free and opensource. P.S: Don't forget to star the repo to show your support
-
oLLM: Lightweight Library for Local LLM Inference
By
–
oLLM is a lightweight Python library for local large-context LLM inference. Run gpt-oss-20B, Qwen3-next-80B, Llama-3.1-8B on ~$200 consumer GPU with just 8GB VRAM. And this is without any quantization – only fp16/bf16 precision. 100% Opensource.
-
RAGLight: Lightweight Open Source RAG Library for Production Systems
By
–
RAGLight: Open Source RAG Library A lightweight Python library for building production-ready RAG systems, featuring LangGraph-powered agent pipelines and multi-provider LLM support with built-in GitHub integration and CLI tools. Check out RAGLight now!