DeepSeek AI unveils SPCT inference scaling technique and hints at R2 model launch, shifting focus from pre-training to post-training optimization.
Zhipu.AI open-sources GLM-4 and GLM-Z1 models, achieving 8x speed over DeepSeek-R1, launches Z.ai, and hints at IPO.
Kwai AI's SRPO matches DeepSeek-R1-Zero performance in math and code using only 10% training steps, solving GRPO's cross-domain and efficiency bottlenecks.
DeepSeek-Prover-V2, an open-source LLM, achieves state-of-the-art 88.9% on MiniF2F theorem proving benchmark using recursive proof search and reinforcement learning.
DeepSeek-V3 paper reveals hardware-aware co-design to slash LLM training costs, offering a blueprint for scalable AI with limited hardware resources.
Stanford, Princeton, and Adobe developed LSSVWM using State-Space Models to enable long-term memory in video AI, overcoming quadratic attention limits.
Researchers from PSU, Duke, and partners introduced Automated Failure Attribution for multi-agent systems, benchmark dataset Who&When, published at ICML 2025.
MIT's SEAL framework enables LLMs to autonomously update weights via self-generated data and reinforcement learning, marking concrete progress toward self-improving AI.
ByteDance unveils Astra, a dual-brain navigation system that splits robot intelligence to solve complex indoor navigation challenges.
Researchers unveil 'Who&When' benchmark to automate failure attribution in LLM multi-agent systems, identifying responsible agents and failure timings. Accepted at ICML 2025.
Learn a step-by-step framework to conceptualize Python fundamentals by defining concepts in your own words, answering key questions, and using visual aids—plus a free cheat sheet and interactive quiz.
Learn to install and configure Codex CLI, use Plan mode for safe reviews, and refine Python code via iterative prompting in the terminal.
Learn unittest fundamentals: structuring tests with TestCase, assertion methods, conditional skipping, parameterization with subtests, and test fixtures. Includes code examples and best practices.
Learn to use Python's unittest framework: TestCase, assert methods, TestSuites, fixtures, and command-line execution for effective automated testing.
Learn how ChatterBot learns from conversations, selects replies via similarity, and integrates local LLMs for richer responses. Step-by-step guide.
Discover Python 3.13's modern REPL with revamped help, multiline editing, smart pasting, and an integrated history browser. Enhance your coding workflow today.
Learn the four types of AI coding agents—IDE, terminal, PR, and cloud—and how to choose the right workflow for your development tasks. Understand the agentic loop and tradeoffs.
Learn Python data analysis workflow: cleansing with pandas, detecting outliers/typos, and regression to find relationships.
Learn the Factory Method pattern in Python: understand its components, when to apply it, and implement a reusable solution to separate object creation from use.
Explore how marimo pair brings AI agent skills into data science notebooks for collaborative pair programming, data wrangling, and research assistance.