One last week of madness in AI: – Runway Game Worlds
– Grok 2.5 Open Source
– Qwen Image Editing Model
– ElevenLabs: from video to music
– Highsfield: from product to video
– NanoBanana debuts on LMArena
– Google AI launches Agentic Mode
LLMS
-
Weekly AI Roundup: Runway, Grok, Qwen, ElevenLabs and More
By
–
-
Stop Using Others’ Prompts, Apply First Principles Thinking
By
–
9/ Stop using other people’s prompts. Start using First Principles Thinking to build your own. You’ll get sharper results. And you’ll understand the output better too.
-
First Principles Thinking in AI prompts for better responses
By
–
8/ You can also use First Principles Thinking in the prompt itself. Tell the AI to think that way. It shifts the whole response. You’ll get structure, clarity, and intent – built from the ground up.
-
Vague prompts yield average ChatGPT newsletter results
By
–
5/ Example: Want ChatGPT to write a newsletter? Most people would ask:
“Write me a good newsletter.” That’s vague. GPT fills in the blanks with generic assumptions. Results = average. -

French Students Open-Source LFM2 Language Model Finetune
By
–
Really impressed by the French finetune of LFM2 made by two students. They created a solid post-training pipeline (FFT + merging) and open-sourced all the code and data. Amazing work by Sinoué Gad and Maxence Lasbordes!
-
LLM Hallucinations and User Perception of Model Improvements
By
–
Human’s hallucination problem with LLMs. “Model got better”, “model got worse” when nothing major changed. GPT-5 launch reception being a great example.
-
Reasoning Models Outperform: Inference-Time Compute Advantage
By
–
Key insights:
Reasoning > non-reasoning models. Better living through inference-time compute!
Smaller open-weights models struggle without fine-tuning or other post-training optimization
SnorkelWordle is a strong signal for evaluating reasoning, especially in smaller models -
Expert-Curated Datasets Improve Small Model Reasoning Skills
By
–
Want to improve a smaller model’s reasoning skills? Our DaaS team is building expert-curated datasets for reasoning, function calling, and more. Let’s talk!
-
SnorkleWordle Benchmark Evaluates LLM Performance Rare Words
By
–
Word games yield valuable insights when evaluating LLMs. We built the SnorkleWordle benchmark to test models on 100 rare English words—and the results are
