Dedicated to sharing and discussing big data and AI technology applications, aiming to empower a million data scientists. Regularly hosts live tech talks and curates articles on big data, recommendation/search algorithms, advertising algorithms, NLP, intelligent risk control, autonomous driving, and machine learning/deep learning.
Anthropic has extended Claude Fable 5’s free access until July 12 and revealed a cost‑saving strategy that pairs Fable 5 as a high‑level planner with the cheaper Sonnet 5 worker, delivering up to 2.5× lower expenses and up to three‑fold faster execution on benchmarks such as SWE‑bench Pro and BrowseComp.
The article dissects the concept of an Agent Harness— the full software infrastructure that wraps large language models—covering its definition, three engineering layers, twelve essential components, step‑by‑step execution loops, framework implementations, and key design decisions that determine whether an AI agent succeeds in production.
Palantir CEO Alex Karp argues that the AI era splits enterprises into those that embed AI‑enhanced infrastructure and those that don’t, warning that relying on generic tools or “parasitic” software leaves companies without a sustainable competitive edge and risks being hollowed out.
The keynote explains how agent architectures have evolved from Prompt, Context, and Harness Engineering to the new Coordination and Symbiosis Engineering paradigms, and how JiuwenSwarm’s AgentOS tackles the enterprise‑level challenges of multi‑agent collaboration, mission‑critical workflows, and large‑scale production deployment.
This article explains how to identify and close the information gap between prompts, skills, and context when working with Claude Fable 5, offering a step‑by‑step workflow—including blind‑spot scanning, brainstorming, interview‑style prompting, reference handling, implementation planning, and post‑release testing—to turn unknowns into actionable knowledge.
Xiaohongshu transformed its data platform from a simple ClickHouse‑based ad‑hoc analysis system to a Lambda‑style architecture and finally to a lakehouse with incremental compute, cutting architecture complexity, resource and development costs by one‑third while delivering second‑level queries over petabyte‑scale data.
The talk presents a three‑layer architecture for enterprise Data Agents, explains how StarRocks‑based AI‑native real‑time data foundations, Semantic View, and context layers work together, and outlines practical checks and feedback loops to address metric, permission, latency, and governance bottlenecks.
This article presents a comprehensive technical analysis of multimodal GraphRAG, covering document‑intelligent parsing pipelines, multimodal graph index construction, knowledge‑graph‑enhanced chunk linking, various multimodal RAG approaches, their trade‑offs, benchmark results, and future research directions.
Although the Caveman plugin claims up to 65% token reduction, independent testing shows real‑world coding sessions only save 4‑10% and that aggressive input compression can actually increase costs by up to 2.7×, because token consumption is dominated by code generation, file reads, and multi‑step Agentic workflows; the article dissects benchmarks, Uber’s budget crisis, and the practical limits of prompt compression.
The article defines Agent Harness as the full software infrastructure that wraps LLMs to enable stateful, tool‑using agents, breaks it down into twelve concrete components, compares implementations from Anthropic, OpenAI, LangChain and others, and outlines key engineering decisions that affect performance, safety and scalability.
