What if we apply RL+Reasoning into PRMs? StepWiser: Stepwise Generative Judges for Wiser Reasoning STEPWISER trains a CoT-based judge that checks whether each “chunk-of-thought” helps solve the problem, producing better step labels and stronger search than discriminative PRMs
RESEARCH
-
AI Technique: realistic fake video from last frame of real video
By
–
It's AI, just in case. It's a fairly simple technique that consists of taking a real video, taking just the last frame and generating a wtf video afterwards. It gives a true illusion of reality, but it never happened.
-

Google Engineer Releases Free Agentic Design Patterns Book
By
–
This Google engineer just released a 424-page free book on Agentic Design Patterns. Covers advanced prompt engineering, multi-agent frameworks, RAG, agent tool use and MCP. 100% free with practical code examples.
-
Core Building Blocks of Large Language Models Explained
By
–
Core Building Blocks of LLMs Prompt Engineering → Zero/Few-shot, Role-based, System Messages, RAG Integration, Chain of Thought Model Architecture → Layers, Residual Connections, Positional Encoding, Attention Training Data → Web Text, Code Repos, Academic
-

AI Development Expected to Drive Unimaginable Discoveries in 1-3 Years
By
–
If AI development continues over the next 1-3 years (which Amodei expects), we will see unimaginable discoveries. Everything will change.
— Chubby♨️ (@kimmonismus) 6 septembre 2025
pic.twitter.com/eMboXYWvEW https://t.co/yWaipeYMIQIf AI development continues over the next 1-3 years (which Amodei expects), we will see unimaginable discoveries. Everything will change.
-
140th LLM Release Covered on Technology Blog
By
–
Just noticed this it the 140th LLM release I've written about in my blog!
-
Kimi K-2.1: Moonshot’s Upgraded Open Weights Model
By
–
My notes on Kimi-K2-Instruct-0905, aka Kimi K-2.1 – an incremental improvement on Moonshot's previous trillion parameter open weights model, now with twice the context length (256k up from 128k)
-

OpenAI’s o1 Release Strategy: Why Holding Back Reasoning Models Mattered
By
–
In retrospect it is surprising that OpenAI released o1-preview. As soon as they showed off reasoning, everyone copied it immediately. And if they had held off releasing a reasoning/planning model until o3 (& called that GPT-5) it would have been a startling leap in AI abilities.
-
OpenAI’s Journey from Open Research to Closed Commercial AI
By
–
> be openAI, circa 2017
> do lots of interesting research
> all open source > no one cares that much
> invent chatGPT in 2022
> get too busy to do open research
> everyone gets mad
> wheres the open research
> more like closedAI
> three years pass. now 2025
> finally release an -
CRISPR Beta Cell Editing Eliminates Immunosuppression Need Type 1 Diabetes
By
–
On the jump forward of CRISPR genome editing beta islet cells, preempting the need for immunosuppression, for a one-shot treatment of autoimmune (Type 1) diabetes https://
nejm.org/doi/full/10.10
56/NEJMe2507578
… https://
nature.com/articles/d4158
6-025-02802-5
…
