Deep think != Deep Research, the latter uses flash and pro
LLMS
-
AI Model News: M2.7, Music 2.6 and TTS
By
–
> M2.7 is a 230B MoE model that partially trained itself. It scores 56.22% on SWE-Bench Pro and is priced at $0.30 per 1M input tokens. > Music 2.6 brings cover generation and BPM controls. > TTS covers 40 languages with seven emotional registers and real-time low-latency
-
AI Models Have Preferred Names: Marcus Chen, Aldric, Kira, Mara Vance
By
–
All of the AI models have preferred names. If you asked Claude 4.5 for a software developer, you are going to get Marcus Chen. Wizards are mostly named Aldric. Space pilots are Kira from Claude, Mara Vance from GPT-5.2. I guess LinkedIn Bros are Kai now. https://
seehuhn.de/blog/ai-names/ -
LatentUM: AI Model Processes Images Text Actions Simultaneously
By
–
What if an AI could think in pictures and words simultaneously, without the usual translation lag? Researchers from Shanghai Jiao Tong U, Tsinghua U, and UCSD present LatentUM. They built a single model that processes images, text, and actions all in one shared "semantic
-
Congratulations for 16 billion tokens per minute
By
–
16B tokens per minute is insane number! Congrats
-
OpenHarness: Open-Source Agent Harness for AI Systems
By
–
Lightweight agent harness for building AI agents! OpenHarness is an open-source agent harness that gives LLMs tools, memory, permissions, and coordination. Here's the concept. The model provides intelligence. But to be a working agent, it needs tools to interact with the world,
-
Freed compute capacity suggests GPT-5.5 will transform the conversation
By
–
If image-2 is what the freed up compute looks like, gpt-5.5 is going to be a different conversation entirely.
-
GPT 5.5 Release Expected Thursday, List Updates Needed
By
–
Ask again after thursday when gpt 5.5 drops. That list might need updating.
-
Claude Code Reverse-Engineered: Only 1.6% Is Actually AI Model
By
–
Researchers reverse-engineered Claude Code. Only 1.6% is actually AI. The other 98.4% is infrastructure around the model. Permission gates, tool routing, context compaction, session recovery. The model just reasons. Everything else runs the show. The core loop is a plain
-
Engineering Ethics: When Not to Build with LLMs
By
–
We’re entering a phase where “can we build it?” matters less than “should we build it this way?” Energy, cost, and latency are becoming very important societal decisions. LLMs are powerful, but they’re not always the right abstraction. Good (AI) engineering is knowing when not