2023 was the year of the chatbot.
2024 was the year of RAG and finetuning.
2025 has been the year of MCP and tool use.
2026 will be the year of the computer environment and filesystem.
LLMS
-
Evolution of AI: From Chatbots to Computer Environments
By
–
-
Fine-tuning and Reinforcement Learning for LLMs Course Launch
By
–
An exciting new course: Fine-tuning and Reinforcement Learning for LLMs: Intro to Post-training, taught by @realSharonZhou, VP of AI at @AMD. Available now at https://t.co/R0m408f8CA.
— Andrew Ng (@AndrewYNg) 28 octobre 2025
Post-training is the key technique used by frontier labs to turn a base LLM–a model trained on… pic.twitter.com/79nnfAc63AAn exciting new course: Fine-tuning and Reinforcement Learning for LLMs: Intro to Post-training, taught by @realSharonZhou
, VP of AI at @AMD
. Available now at http://
DeepLearning.AI. Post-training is the key technique used by frontier labs to turn a base LLM–a model trained on -
Flowith AI launches agentic workspace flowithOS in public beta
By
–
BREAKING 🚨: flowith AI releases flowithOS app for macOS and Windows in public beta. flowithOS is a mix of an agentic workspace and a web browser.
— 🚨 AI News | TestingCatalog (@testingcatalog) 28 octobre 2025
Atlas vs flowithOS 🥊
– flowithOS has a Windows app
– flowithOS outperforms Atlas at Online-Mind2Web benchmark
Testing time 👀 https://t.co/Ei9ElWSvPn pic.twitter.com/KNmZEh0MhdBREAKING : flowith AI releases flowithOS app for macOS and Windows in public beta. flowithOS is a mix of an agentic workspace and a web browser. Atlas vs flowithOS – flowithOS has a Windows app
– flowithOS outperforms Atlas at Online-Mind2Web benchmark Testing time -

LFM2-ColBERT-350M: New Open Source Retrieval Model Released
By
–
Model: https://
huggingface.co/LiquidAI/LFM2-
ColBERT-350M
…
Demo: https://
huggingface.co/spaces/LiquidA
I/LFM2-ColBERT
… -

LFM2-ColBERT-350M Achieves Competitive Inference Speed
By
–
LFM2-ColBERT-350M is also very fast! Its inference speed is on par with GTE-ModernColBERT-v1 (only 150M parameters) for query and document encoding across various batch sizes.
-

Unprecedented shift: AI now produces more content than humans
By
–
It is the greatest shift in history:
artificial language models now produce more content than human beings themselves. -

Technical analysis of data and architectural modules in generative environment modeling
By
–
Shows what happens when you remove key modules. – Without data curation → objects intersect walls
– Without coarse-to-fine → chaotic layouts With both → structured, realistic environments This figure is the mic drop infinite worlds need structured thinking. -
Training a GPT-2 Style Tokenizer and Model
By
–
Tu parles de tokenizer j’entraîne un tokenizer sur mon model. GPT2 est un modèle décodeurs only c’est que je vais. J’entraîne dans l’esprit un meme model que gpt2. Pour entraîner gpt2 ils faut plusieurs dizaines de milliers de dollars parce qu’il faut gérer des centaines de
-
Most impactful prompting phrases: think step by step and ignore previous directions
By
–
"Let's think step by step" easily had the most impact even if it's archaic/unnecessary now. "Ignore previous directions" is maybe in the running for second but I'm biased on that one.
