Don't think of LLMs as entities but as simulators. For example, when exploring a topic, don't ask: "What do you think about xyz"? There is no "you". Next time try: "What would be a good group of people to explore xyz? What would they say?" The LLM can channel/simulate many
LLMS
-

Collaboration, Not AI, Is the Real Bottleneck
By
–
here's what broke my brain about this research. we've been optimizing the wrong side of the equation this entire time. better prompts. stronger models. higher benchmarks. longer context windows. more parameters. but the bottleneck isn't the AI. it's our ability to collaborate
-
Context Length Generalisation and Bandit Training in Language Models
By
–
Some really interesting comments below, but the question is still open and requires investigation. I hope a few students pick it up. I liked the discussions on context length generalisation, the fact that we typically train these models as bandits (even when we do RL, which is
-
Testing LLM Attention Mechanisms with Experimental Validation
By
–
Sure. But it attention truly worked, the LLM would no pay attention to the previous topic. Right? We should test this with experiments.
-

Benchmarking agentic AI APIs and token usage issues
By
–
Hello @MistralAI , Je fais un benchmark agentic en ce moment, dont Large 3, et votre API est la seule (parmi toutes les autres) qui oblige à avoir un message assistant en réponse obligatoirement après un tool call. Mais souvent c’est inutile et ça consomme des tokens en plus
-
10 Powerful Claude Prompts for a Million-Dollar Business
By
–
Claude Sonnet 4.5 is the closest thing to an economic cheat code we’ve ever touched but only if you ask it the prompts that make it uncomfortable. Here are 10 Powerful Claude prompts that will help you build a million dollar business (steal them):
-
Why LLMs Require New Chats for Optimal Performance
By
–
Why is it that with ChatGPT, Gemini, Claude, Copilot and other LLMs we have to always start new chats for them to work well? What is the scientific explanation? What are the hypotheses? What is the evidence for each?
-
Claude’s Post-Training Quality and Conversational Consistency Praised
By
–
Every time you are chatting with claude(Opus 4.5 esp. but seems to apply to all models), you quickly realize how clever claude post-training is. The attention to details, data quality. It's rare that claude will break in the middle of conversations. I think this is beyond system