GLM 4.6 will continue to exist and be useful for powering coding agents even if every single AI lab goes bust and vanishes
LLMS
-

SnorkelSpatial Benchmark Tests LLM Spatial Reasoning Abilities
By
–
New benchmark drop SnorkelSpatial tests how well LLMs can think in space, following text-based moves and rotations in a 2D world.
-
New Annotations Feature Released for Google AI Studio
By
–
The annotations feature for @GoogleAIStudio Built was also released! For example, you can now precisely instruct Gemini on which animations to add to your app. https://t.co/AHbrkPuBnQ pic.twitter.com/LDxjYsQsBk
— 🚨 AI News | TestingCatalog (@testingcatalog) 23 octobre 2025The annotations feature for @GoogleAIStudio Built was also released! For example, you can now precisely instruct Gemini on which animations to add to your app.
-
Stitch Application Upgraded with Gemini 2.5 Pro
By
–
Stitch is now powered by max-tuned Gemini 2.5 Pro, and it is killing it. There is a visible difference from what Stitch was capable of before. https://t.co/CNSuLzOAdX pic.twitter.com/R1awz7X210
— 🚨 AI News | TestingCatalog (@testingcatalog) 23 octobre 2025Stitch is now powered by max-tuned Gemini 2.5 Pro, and it is killing it. There is a visible difference from what Stitch was capable of before.
-
Claude Opus rate limits on $20 subscription plan
By
–
I can send about 15 messages to Opus on the $20 plan before I get locked out for 4 hours.
-

Meta’s Soft Tokens Enable LLMs to Invent Recursive Representations
By
–
This paper from Meta about "Soft Tokens" in RL is interesting; it allows LLMs to invent their own non-discrete (recursive) representations in order to solve problems better… Results are mixed though: it's only a few percent better on GSM8k from pass@4 onwards, and pass@32 just
-
AI Dev 25 NYC: Agentic Architecture and Production Systems
By
–
The full agenda for AI Dev 25 x NYC is ready. Developers from Google, AWS, Vercel, Groq, Mistral AI, SAP, and other exciting companies will share what they've learned building production AI systems. Here's what we'll cover: Agentic Architecture: When orchestration frameworks
-
Google developing Genie 3 AI world model for user-generated exploration
By
–
BREAKING 🚨: Google started working on a Genie 3 experiment where users will be able to generate their own worlds to explore.
— 🚨 AI News | TestingCatalog (@testingcatalog) 23 octobre 2025
Genie 3 is an AI World Model from Google announced earlier this year. https://t.co/4122r1EqhO pic.twitter.com/Mh24PcsYgvBREAKING : Google started working on a Genie 3 experiment where users will be able to generate their own worlds to explore. Genie 3 is an AI World Model from Google announced earlier this year. https://
x.com/synthwavedd/st
/synthwavedd/status/1981382737698132474
… -
NotebookLM to Introduce Deep Research Feature for Data Sourcing
By
–
BREAKING 🚨: NotebookLM is about to release Deep Research for data sourcing on NotebookLM soon, as a new announcement label has been added.
— 🚨 AI News | TestingCatalog (@testingcatalog) 23 octobre 2025
Here is an early preview 👀 pic.twitter.com/ovT2QhZNBABREAKING : NotebookLM is about to release Deep Research for data sourcing on NotebookLM soon, as a new announcement label has been added. Here is an early preview