Thanks @xeophon and @natolambert for your kind words on @interconnectsai LFM2.5-1.2B-Thinking is even more impressive in terms of instruction following. Please check it out!
LLMS
-
Technical challenges in VLM spatial reasoning and 3D scene graphs
By
–
Thanks. Another missing piece for now is VLMs having the spatial reasoning to manipulate 3d scene graphs effectively. But that hill is actively being climbed by the labs and is a matter of time.
-
Codex AI Model Enables High-Velocity Development in Open Source Projects
By
–
100% yes – infact the OSS repo is probably the biggest public case study for using codex at scale across teams and tech stacks surreal to see it enabling teams deliver at such insane rate with confidence
-
xAI Announces Grok Imagine 1.0 for Enhanced Video and Audio Generation
By
–
xAI announced Grok Imagine 1.0 with 10-second videos and improved audio.
— 🚨 AI News | TestingCatalog (@testingcatalog) 2 février 2026
Upgraded video generation is now available on Grok apps and APIs. Generations are also quite different from other top models.
"Cyberpunk robot test" 👀 https://t.co/c79zc7CFhQ pic.twitter.com/G8EbWfR8ofxAI announced Grok Imagine 1.0 with 10-second videos and improved audio. Upgraded video generation is now available on Grok apps and APIs. Generations are also quite different from other top models. "Cyberpunk robot test"
-
Practical Tips for Tuning AI Agent Behavior and Workflows
By
–
some recommendations on making it tuned towards how you drive it: 1. use /init to create an agents md file
2. everytime codex does something wrong ask it to update its agents md to do/ not do X
3. use plan mode to plan and write the plan in a plans md file
4. ask codex to -
User request for expanded multimodal input in Genie AI
By
–
Very doable! Tho I wish genie would just accept video input and multi view images too https://t.co/Z05LOz9lVV
— Bilawal Sidhu (@bilawalsidhu) 2 février 2026Very doable! Tho I wish genie would just accept video input and multi view images too
-
Anticipation for Grok to be connected to lists for hyper usefulness
By
–
I can't wait for @grok to get hooked up to lists. When it does these will become hyper useful.
-
User prefers Blevlabs over Grok due to superior quality
By
–
Yeah, but Grok isn't nearly as good as @blevlabs so I'll just keep using that.
-
Grok connected to lists could bring similar advantages
By
–
Of course, if Grok got hooked up to lists we would get most of the same advantages.
-
LLM Brain Rot: How AI Models Degrade from Low-Quality Data
By
–
I feel like I am actively getting dumber. LLMs get brain rot and it is measurable: https://
llm-brain-rot.github.io
"continual exposure to junk web text induces lasting cognitive decline in large language models (LLMs)" why shouldn't the same be true for brains.
