"Deep Researcher with Test-Time Diffusion" This paper treats report writing as an iterative retrieval‑augmented diffusion process that can be enhanced by component‑wise self‑evolution. This demonstrates SoTA on multi‑hop search‑and‑reasoning benchmarks.
GENERATIVE AI
-
AI Model Comparison App Built with MagicPath
By
–
Comparing models is hard, so I built a little app where you can compare any model you want.
— Pietro Schirano (@skirano) 23 juillet 2025
All you need is the @OpenRouterAI API.
You can compare up to four models and even have a dedicated system prompt for each.
Entirely built on MagicPath with one prompt. pic.twitter.com/frsokxeWrTComparing models is hard, so I built a little app where you can compare any model you want. All you need is the @openrouter API. You can compare up to four models and even have a dedicated system prompt for each. Entirely built on MagicPath with one prompt.
-
AI Tool Enables Interactive Climate System Exploration
By
–
Introducing: Climate in a Bottle Our new AI-powered tool lets users explore the climate system interactively. Set the season, ocean temperature, time of day—and generate high-resolution climate states instantly.
-
VEO 3 Logo Transformation: JSON Prompts Improve Video Outputs
By
–
La transformation de logos dans VEO 3
— VISION IA (@vision_ia) 23 juillet 2025
Les gens ont découvert qu'utiliser des prompt au format JSON améliore grandement les sorties vidéos https://t.co/MkMmjuVb7L pic.twitter.com/TATUAyLD46La transformation de logos dans VEO 3 Les gens ont découvert qu'utiliser des prompt au format JSON améliore grandement les sorties vidéos
-
Google AI Model Restores Lost Latin Inscriptions with 44% Accuracy Boost
By
–
Neat example of AI in the humanities. A Google model trained on Latin text fills in lost parts of Latin inscriptions & identifies related texts Historians increased their accuracy by 44% when working with the AI (Though AI alone beats historians, historian + AI was usually best)
-
Author Removes Em-Dashes Due to AI Concerns
By
–
A publisher friend just told me an author removed all em-dashes from her manuscript because of AI. I follow several linguists on TikTok and Instagram, plus have a lot of friends in the writing space. We’re just barely scratching the surface on the impact of AI on everyday
-
Hugging Face Benchmark: Vision LLMs Long Video Input Performance
By
–
Interesting new benchmark from Hugging Face testing how well vision LLMs can handle long video inputs (generally after they've been split into many thousands of images) – my notes here: https://
simonwillison.net/2025/Jul/23/ti
mescope/
… -
Scientific Papers Enhanced With LLM-Assisted Demos for Accessibility
By
–
Aside from everything else interesting about this paper, I appreciate that more scientific papers (aided by LLM help?) are now including little demos and experiments to help non-specialists get the points they are making. (And no, you cannot identify the hidden signals)
-
AI-first media maker hailed as top Canva alternative
By
–
Canva – take note. This is probably the best alternative I've seen that's designed as an 'AI-first' media maker.
— AI Breakfast (@AiBreakfast) 23 juillet 2025
Very impressive. https://t.co/7q2CFZMfrMCanva – take note. This is probably the best alternative I've seen that's designed as an 'AI-first' media maker. Very impressive.
-
When Synthetic Data Diverges from Reality: Implications Explored
By
–
What happens when data breaks from reality?
— SAS Software (@SASsoftware) 23 juillet 2025
In The Shape of Synthetic Data, Dietmar Offenhuber joins Pondering AI host Kimberly Nevala to explore the speculative nature of synthetic data, the materiality of information, and why meaning isn't always digital—or even statistical.… pic.twitter.com/YpYNDIbIofWhat happens when data breaks from reality? In The Shape of Synthetic Data, Dietmar Offenhuber joins Pondering AI host Kimberly Nevala to explore the speculative nature of synthetic data, the materiality of information, and why meaning isn't always digital—or even statistical.