brb i’m gonna use claude code with opus 4.5 to refactor ubuntu in react native
LLMS
-

Instructed Retriever: Enterprise Agents With System-Level Reasoning
By
–
Reliable enterprise agents require system-level reasoning when retrieving across heterogeneous knowledge sources. Traditional RAG often fails to consistently follow instructions, schemas, and constraints end to end. That’s why we’re presenting Instructed Retriever, a new
-
Clarification on Opus 4.5 Assessment and Pod Commentary
By
–
All of the following are true: I came out sounding too critical in that moment on the pod and accidentally created a misrepresenting sound bite, I was positively surprised by Opus 4.5 over the break, and my overall change of mind overall is less dramatic than naively implied.
-

GPT-5.2-Codex-Max Testing Reports Surface
By
–
OpenAI might be testing GPT-5.2-Codex-Max already with some users as there are multiple reports where Codex claims to be upgraded. Mine is still 5.1, what did you get? h/t @Haha58208745
-
Knowledge Assistant transforms unstructured data into searchable intelligence
By
–
[Demo] Explore the Knowledge Assistant for unstructured data 📑
— Databricks (@databricks) 6 janvier 2026
See how unstructured documents can be turned into a searchable knowledge layer. The assistant ingests and parses documents from Unity Catalog files or vector search indexes, then uses a RAG chatbot to return answers… pic.twitter.com/pnY44sCGm0[Demo] Explore the Knowledge Assistant for unstructured data See how unstructured documents can be turned into a searchable knowledge layer. The assistant ingests and parses documents from Unity Catalog files or vector search indexes, then uses a RAG chatbot to return answers
-
Building High-Quality Deep Research Agents While Reducing Token Waste
By
–
🎙️ How do you make a high quality deep research agent without burning thousands of tokens?@tavilyai's Dean Sacoransky joins our host, @yuvalinthedeep on log-driven iteration, cutting wasted tokens, and improving quality section by section.
— AI21 Labs (@AI21Labs) 6 janvier 2026
Watch it here:… pic.twitter.com/sQDWlvV0xRHow do you make a high quality deep research agent without burning thousands of tokens? @tavilyai
's Dean Sacoransky joins our host, @yuvalinthedeep on log-driven iteration, cutting wasted tokens, and improving quality section by section. Watch it here: -

ScaledML Returns After 6 Years with ML Industry Foresight
By
–
ScaledML is back! After a 6 year hiatus. Consistently 2-3 years ahead of where ML will be. Examples of foresight at SML:
– OpenAI in 2016,2017,2019 announced GPT-2 & RL efforts
– Turing award for Deep Learning announced by Turing award winner on morning of award
– Groq chip -
Normal LLMs store knowledge in high-precision weights causing huge memory
By
–
This is insane, here’s how this works:
— AI Breakfast (@AiBreakfast) 5 janvier 2026
Normal LLMs (like Grok or GPT variants) store their "kn0wledge" in billions of numbers (weights)
Each weight is usually a high-precision decimal (like 3.14159), stored using 16 bits per weight. This makes the model huge in memory and… https://t.co/P2W0pNNqxqThis is insane, here’s how this works: Normal LLMs (like Grok or GPT variants) store their "kn0wledge" in billions of numbers (weights) Each weight is usually a high-precision decimal (like 3.14159), stored using 16 bits per weight. This makes the model huge in memory and
-

Sakana AI Agent Wins Optimization Contest Using Test-Time Inference
By
–
So proud of Team Sakana AI for pulling this off! We managed to get an agent to rank #1 in a difficult heuristic optimization contest. We did this by leaning heavily into test-time inference using a mix of frontier models. The agent spent about $1,300 in credits to autonomously
-
Gemini 3 Outperforms Frontend Plugin in Performance Comparison
By
–
Gemini 3 outperforms the front end plugin for me 8 times out of 10