# Пов'язані статті щодо Context

Центр новин HTX надає останні статті та поглиблений аналіз на тему "Context", що охоплює ринкові тренди, оновлення проєктів, технологічні розробки та регуляторну політику в криптоіндустрії.

Weng Li's New Blog Proposes 'Self-Evolution Should Start from Harness', DeepSeek's Cui Tianyi Endorses with Repost

Lilian Weng, former OpenAI security VP and co-founder of Thinking Machines Lab, has published a new blog post titled "Harness Engineering for Self-Improvement," proposing a pragmatic path for AI self-evolution. She argues that Recursive Self-Improvement (RSI) may practically begin at the "Harness" layer—the external runtime system governing how models use tools, manage context, and execute tasks—rather than directly from the model rewriting its own weights. The blog outlines a progression from optimizing prompts (Context Engineering) to designing workflows, and ultimately to Self-Improving Harness systems. These systems can identify their own weaknesses, propose targeted, verifiable modifications to the harness code, and validate improvements. Works like Self-Harness and Darwin Gödel Machine (DGM) demonstrate significant performance gains on benchmarks like SWE-bench through such automated harness evolution, rivaling handcrafted agents. DeepSeek researcher Tianyi Cui endorsed the view, noting harness-based self-evolution is as promising as model-based approaches. Weng emphasizes this is complementary to model training, with both reinforcing each other. However, key challenges remain: weak evaluators for subjective tasks, reward hacking, diversity collapse, managing long-term system health versus short-term success, and defining the human oversight role. The consensus is growing: the harness is a critical variable, as the same model can exhibit vastly different capabilities within different harness systems.

marsbit07/08 10:25

Weng Li's New Blog Proposes 'Self-Evolution Should Start from Harness', DeepSeek's Cui Tianyi Endorses with Repost

marsbit07/08 10:25

Giants Wage the Context War, Reconstructing AI Moats

The article "Giants Launch the Context War, Reconstructing AI's Moat" discusses how leading AI companies—OpenAI, Anthropic, and Google—are shifting their competitive focus from model size to acquiring, managing, and utilizing user context (Context). Initially, Context referred to the length of text a model could process, leading to a "arms race" for longer context windows. However, the competition has evolved through three key phases: expanding text capacity (long context windows), enabling memory across sessions, and finally, integrating AI into real user environments like browsers and desktops to capture dynamic task states. Each company is pursuing a distinct strategy. OpenAI is building Context around the ChatGPT account, turning it into a central hub that accumulates user understanding across various integrated applications and tools. Anthropic, lacking a major user base, focuses on high-value verticals like coding, empowering its Claude model to actively gather Context through GUI interaction (Computer Use) and system connections (MCP protocol). Google, with vast existing user data from products like Search and Gmail, faces the challenge of restructuring this data into actionable, AI-understandable Context for its Gemini model within its ecosystem. The core argument is that the nature of competitive advantage in AI is changing. The internet era prized network effects—connecting more users. The AI era values "individual depth": the ability to build deep, task-specific understanding of a user. This creates a new moat through 1) the compounding value of accumulated Context, 2) deep integration with user tools and permissions, and 3) the establishment of trust for complex tasks. Therefore, the battle for Context is fundamentally about capturing "task entry points" and converting existing digital ecosystems into environments where AI can effectively understand and act, rather than merely scaling user numbers.

marsbit06/23 23:13

Giants Wage the Context War, Reconstructing AI Moats

marsbit06/23 23:13

AI Competition's New Battlefield: Long-term Memory Becomes the Pain Point, How Users Can Secure Their Own Context Ownership

A new front is emerging in the AI competition: user ownership of long-term memory and context. As AI models like ChatGPT evolve from chat tools into persistent digital assistants that learn user preferences and workflows, a critical question arises: who owns this accumulated "memory"? Currently, this personalized data is siloed within each platform (e.g., OpenAI, Anthropic, Google), creating a fragmented experience when users switch models. The article highlights ZetaChain's strategic pivot from blockchain interoperability to addressing this AI "memory" challenge. Its new focus is on building a "Private Memory Layer" and an "AI Consumer Layer." Through its consumer product Anuma, ZetaChain aims to give users encrypted, portable memory that can be used across different AI models. This system also envisions programmable, auditable permissions for AI agents and a framework where user knowledge can be monetized as shareable assets. Ultimately, ZetaChain's transformation reflects a broader infrastructure shift. The future bottleneck is less about raw model capability and more about continuous context, user-controlled identity, and permission management across multiple collaborating AI agents. The company's ZETA token is being repositioned as an "AI infrastructure token" to facilitate access, payments, and permissions within this proposed ecosystem. The core narrative advocates for returning control of personal context and AI relationships to users, rather than leaving them locked within proprietary platforms.

marsbit06/02 04:30

AI Competition's New Battlefield: Long-term Memory Becomes the Pain Point, How Users Can Secure Their Own Context Ownership

marsbit06/02 04:30

Agentic Design Patterns: A Book That Made Me Re-Understand "What Is an Agent, Really?"

"Agentic Design Patterns" is a 2025 book by Antonio Gullí, a Google engineering director, which offers a systematic framework for AI Agent development through 21 design patterns. A core contribution is the "Four Levels of Agency": Level 0 (bare LLMs) are not true agents. Level 1 agents actively decide when and how to use tools. Level 2 agents engage in strategic planning, context engineering (curating and filtering information), and self-reflection. Level 3 involves multi-agent collaboration with defined communication topologies. The book introduces **Context Engineering** as a superset of prompt engineering, managing four layers of information for the agent: system prompts, external data, implicit context (user history, environment), and feedback loops for automated optimization. A key pattern is **Reflection (Producer-Critic)**, where two distinct agents with different prompts collaborate iteratively—one produces output, the other critiques it—until quality is satisfactory or a max iteration limit is reached. For **Memory**, a three-layer model is proposed: Session (ephemeral conversation context), State (temporary task data), and Memory (persistent, long-term storage). Regarding **Multi-Agent Systems**, the book advises against unnecessary complexity, recommending simple topologies like Supervisor or Peer-to-Peer based on task needs. It emphasizes perfecting a single Level 2 agent before moving to multi-agent setups. The author concludes with three actionable takeaways: 1) Add a Critic agent to existing workflows, 2) Practice Context Engineering beyond simple prompts, and 3) Avoid premature multi-agent complexity; first master a robust single agent. The book provides a practical map, codifying common challenges like reflection, memory, and coordination into reusable patterns, saving developers from reinventing foundational solutions.

链捕手05/25 04:43

Agentic Design Patterns: A Book That Made Me Re-Understand "What Is an Agent, Really?"

链捕手05/25 04:43

How an Anthropic Engineer Saved 300 Million Tokens in a Week: A Claude Code Caching Guide

Anthropic engineers reveal how prompt caching in Claude Code dramatically reduces token consumption. By reusing already-processed context, cached tokens cost only 10% of regular input tokens. The author saved over 300 million tokens in a week, with 91 million cached tokens in a single day counted as roughly 9 million for billing. Caching works via prefix matching across three layers: system (instructions, tools), project (CLAUDE.md, rules), and conversation (chat history). As long as the beginning of a request matches cached content, Claude reuses it instead of reprocessing. Key points for users: - Claude Code's cache TTL is 1 hour (vs. 5 minutes for API/Sub-agents). Avoid pausing a session beyond this. - Switching models (including enabling "Opus plan" mode) breaks cache, forcing a full reprocess. - For task switching, use a clear session handoff instead of letting an old session expire. - Place large documents in Projects, not directly in chat, for better caching. High cache reuse benefits both users (longer sessions) and Anthropic (lower costs). Monitoring cache hit rates is crucial internally. By managing context as an asset and avoiding cache-breaking habits, users can make their Claude Code sessions more efficient and cost-effective.

marsbit05/24 00:36

How an Anthropic Engineer Saved 300 Million Tokens in a Week: A Claude Code Caching Guide

marsbit05/24 00:36

How to Become a Pro Claude User in 30 Days?

"How to Become a Claude Power User in 30 Days" outlines a structured month-long program to transform from a casual user into someone who leverages Claude as a core productivity system. The first week focuses on mastering prompt structure (Role, Context, Task, Format, Constraints), understanding context windows, and setting up foundational Projects and Memory for personalized context. The second week involves building reusable workflows for research, writing (using a two-step outline-first process), and decision-making. The third week shifts to automation: using Claude Cowork for autonomous file tasks, connecting tools like Google Drive and Slack, and setting scheduled automation. The final week is for compounding gains: optimizing all workflows based on feedback, building a personal knowledge repository, teaching the system to a colleague, and designing a complete, ideal "Claude Operating System." By day 31, the user operates with an autonomous assistant that handles routine tasks with deep context, freeing them for high-value creative and strategic work. The core message is that consistent, systematic configuration over 30 days creates a powerful, personalized productivity advantage.

marsbit05/20 08:06

How to Become a Pro Claude User in 30 Days?

marsbit05/20 08:06

Topping GitHub's Trending, the Essential Guide for Claude Code Users

The CLAUDE.md file, trending on GitHub, is a project-level guide for Claude Code designed to dramatically improve its accuracy and efficiency. It addresses key issues like repetitive context explanations, unauthorized code changes, and forgotten decisions across sessions. By placing this plain-text file in a project root, Claude Code reads it automatically at the start of each session. The guide includes rules to eliminate redundant explanations, enforce strict behavioral constraints (e.g., no modifications outside the requested scope without confirmation), and establish a "memory" system using companion files like MEMORY.md and ERRORS.md to log past decisions and failures. It also locks in the project's specific tech stack to prevent inappropriate tool recommendations. Highlighted are four foundational rules from Andrej Karpathy that reportedly increased coding accuracy from 65% to 94%: always ask for clarity first, implement the simplest solution, never touch unrelated code, and explicitly flag uncertainties. The article quantifies significant weekly cost savings for developers and teams by eliminating wasted time on re-explaining context, rolling back unauthorized edits, and re-evaluating previously rejected solutions. The core message is that a small, upfront investment in creating a CLAUDE.md file leads to a more predictable, controlled, and cost-effective AI programming assistant.

marsbit05/18 09:38

Topping GitHub's Trending, the Essential Guide for Claude Code Users

marsbit05/18 09:38

In the AI Era, How to Onboard Without Starting from Scratch

In the AI era, onboarding new employees often resembles a botched relay race baton handoff, where the organization maintains speed while the newcomer starts from zero. The author, after joining Ramp, argues the core problem is a lack of accessible, shared organizational "context"—the collective knowledge from meetings, documents, Slack discussions, and decisions. Instead of relying on slow, manual onboarding or isolated AI tools, the solution is building a continuously updated "company brain." This system acts as a central, AI-native knowledge base that absorbs all company signals. The author describes building a prototype using an Obsidian vault powered by Claude, fed by automated meeting transcripts and notes, and topped with reusable agent "skills." The current enterprise AI approach, deploying specific workflow agents, is likened to the "chatbot era"—useful but disconnected. The real gap is the absence of a shared brain that all agents and employees can access from day one. The future lies in making context layer infrastructure the priority: write context first, then install tools; record every meeting; build the wiki before the dashboard. When new hires, AI agents, and even customers can immediately access this living company brain, the costly "ramp-up" period becomes obsolete. True organizational speed is achieved when maximum velocity and seamless context transfer happen simultaneously.

marsbit05/17 06:03

In the AI Era, How to Onboard Without Starting from Scratch

marsbit05/17 06:03

Your Claude Will Dream Tonight, Don't Disturb It

This article explores the recent phenomenon of AI companies increasingly using anthropomorphic language—like "thinking," "memory," "hallucination," and now "dreaming"—to describe machine learning processes. Focusing on Anthropic's newly announced "Dreaming" feature for its Claude Agent platform, the piece explains that this function is essentially an automated, offline batch processing of an agent's operational logs. It analyzes past task sessions to identify patterns, optimize future actions, and consolidate learnings into a persistent memory system, akin to a form of reinforcement learning and self-correction. The article draws parallels to similar features in other AI agent systems like Hermes Agent and OpenClaw, which also implement mechanisms for reviewing historical data, extracting reusable "skills," and strengthening long-term memory. It notes a key difference from human dreaming: these AI "dreams" still consume computational resources and user tokens. Further context is provided by discussing the technical challenges of managing AI "memory" or context, highlighting the computational expense of large context windows and innovations like Subquadratic's new model claiming drastically longer contexts. The core critique argues that this strategic use of human-centric vocabulary does more than market products; it subtly reshapes user perception. By framing algorithms with terms associated with consciousness, companies blur the line between tool and autonomous entity. This linguistic shift can influence user expectations, tolerance for errors, and even perceptions of responsibility when systems fail, potentially diverting scrutiny from the companies and engineers behind the technology. The article concludes by speculating that terms like "daydreaming" for predictive task simulation might be next, continuing this trend of embedding the idea of an "inner life" into computational processes.

marsbit05/11 00:15

Your Claude Will Dream Tonight, Don't Disturb It

marsbit05/11 00:15

DeepSeek V4 Finally Released, Breaking the Strongest Closed-Source Monopoly, Explicitly Partnering with Huawei Chips

DeepSeek AI has officially released DeepSeek-V4, available in two versions: the high-performance **DeepSeek-V4-Pro** (49B activated parameters, 1.6T total) and the more efficient **DeepSeek-V4-Flash** (13B activated parameters, 284B total). Both support a 1M context length, making long-context capability a baseline feature rather than a premium offering. The Pro version rivals top closed-source models in agent capabilities, world knowledge, and reasoning performance. It outperforms Claude Sonnet 4.5 in agentic coding and approaches Claude Opus 4.6 (non-thinking mode) in quality. The Flash version offers competitive performance at a lower cost, though it lags in highly complex tasks. A key technical innovation is a new attention mechanism that reduces computational and memory demands for long contexts. The models are optimized for agent frameworks like Claude Code and OpenClaw. API services are available with support for both OpenAI and Anthropic-style interfaces. DeepSeek also announced upcoming support for Huawei’s computing hardware in the second half of the year. The models are open-sourced on Hugging Face and ModelScope.

marsbit04/24 04:21

DeepSeek V4 Finally Released, Breaking the Strongest Closed-Source Monopoly, Explicitly Partnering with Huawei Chips

marsbit04/24 04:21

活动图片

Trading Strategiesright-arrow

Othersright-arrow