All articles
38278 articles
Liang Sheng's Open-Source Super Agent Architecture: Keeping AI Online Long-Term to Work Proactively
Liang Sheng open-sources the DeepSick Operating Prototype super agent architecture, using Router/Worker two-layer decoupling, event-driven design, and tiered inference filtering to turn AI from passively waiting for instructions into a long-term online Agent that proactively identifies problems and executes tasks. This article analyzes its engineering design and cost-control approach.
What DeepSeek Open-Sourced Isn't a Model—It's AI's 'Shut-Up Discipline'
DeepSeek's DSH Super Agent open-source release isn't about model capability but AI work discipline: Router/Worker architecture, subscription passthrough, speculative execution, and Watchdog supervision to solve the human attention bottleneck in multi-task collaboration.
Tech FrontiersLibghostty 1.0 Released: The Terminal Core Library Goes Independent
Libghostty 1.0 is released as Ghostty splits its terminal core into a reusable library. We analyze the engineering rationale, ecosystem impact, and community feedback.
Tech FrontiersPerplexity Refreshes Its Pareto Frontier Every Month: A New High for AI Search Efficiency
Perplexity refreshes its Pareto optimal curve monthly to keep pushing the cost, speed, and quality boundaries of AI search. Here's what this cadence means for competition and developers.
Tech FrontiersAgent Calling Agent: A New Paradigm of Recursive Collaboration with Git as the Hub
Agents will increasingly be called by other Agents, driven by simpler orchestration and centralization around Git as the system record. Recursive loops work when Agents run in code-mode with version-controlled traces, measured by less human intervention.
Tech FrontiersPerplexity Internal Data: AI Agent Usage Grows 7x, 1 in 5 Employees Exceeds 40 Agent-Hours per Week
Perplexity internal data shows employees' median weekly agent-hours from teammates is now 7x September's level, rising every week, with 1 in 5 exceeding 40 hours — equal to an extra full-time hire.
Lean Verified Tokenizers: Solving the RL Training-Inference Mismatch Problem
The Lean Verified Tokenizers project tackles training-inference mismatch in RL training by formally verifying tokenizers and message pipelines with Lean, and prototyping a Rust agent pipeline as an alternative to error-prone Jinja templates.
Tech FrontiersA Tweet Reveals a New Path for Idea Dissemination in AI
Starting from a tweet by Jiajun Li, this article analyzes the social media path of idea dissemination in AI, the issue of idea attribution, and implications for the content ecosystem, exploring the value of connecting fragmented citations with in-depth blogs.
AI Coding Enters the Team Collaboration Era: Interpreting Three Major Moves from Meta, GitHub, and Mistral
Meta acquires Conductor, GitHub Copilot Fusion becomes available, Mistral opens Large 4 weights—three moves show AI coding shifting from code generation to organizing development workflows. This article explains task orchestration, model collaboration, and cost control for developers.
Building a RAG Assistant with Dify: A Practical Guide to Knowledge Bases, Workflows, and Agent Nodes
How should knowledge bases, workflows, and Agent nodes connect in a Dify RAG assistant? We break down the division of labor and wiring logic for a shippable app.
Tech FrontiersThe AI Industry Doesn't Lack Marketing—It Lacks Good Products
AI faces market selloffs and public skepticism—is marketing really the problem? This analysis of market signals and product experience argues the core困境 is incomplete products and a gap between promises and experience; the only way to change perceptions is to ship better products.
Tech FrontiersGenerative UI Streaming in Chat SDK Agents: A Deep Dive into json-render
A deep dive into json-render and Chat SDK's generative UI streaming capabilities: typed component catalogs, streamCard incremental updates, toCard spec validation with platform-limit fallbacks, and real-world deployment on Slack, Teams, and more.
Tech FrontiersHow Do You Stop AI Agent Emails From Sounding Like AI? One Developer's Solution
AI agent emails always sound like AI? A developer shares reright: a rewrite step that validates numbers, names, and links, preventing humanizers from tampering with key info. Test data and integration details inside.
Tech FrontiersStarlink Mobile: SpaceX's Carrier Ambitions and the 5G Deployment Challenge
SpaceX plans to make Starlink Mobile a mainstream carrier, deploying thousands of direct-to-cell satellites and building a terrestrial 5G network to challenge AT&T, Verizon, and T-Mobile. This article analyzes its dual-track strategy, 5G deployment challenges, and the competitive landscape.
Tech FrontiersOpenAI Raises Another $11 Billion, Valuation Hits $1.25 Trillion
OpenAI has raised an additional $11 billion in a new funding round at a valuation of $1.25 trillion. The deal is not yet closed and is expected to be completed in the coming weeks, further consolidating its leading position in AI.
Tech FrontiersChina Accused of Recruiting AI Executive to Monitor Jensen Huang as Chip Rivalry Heats Up
The New York Times reports China's MSS recruited Megaspeed executive Alice Huang to gather U.S. AI chip intelligence and influence Nvidia's chip supply to China.
Tech FrontiersAI Gateway Integrates DeepSeek Search and Fetch Tools, Enabling Web Retrieval with a Single API Key
Vercel AI Gateway adds DeepSeek Search and DeepSeek Fetch tools, letting developers add web search and page fetching to any tool-calling model with one API key, keep tool config across model providers, and get multi-source fallback with per-request billing.
Tech Frontiers@ai-sdk/vercel 2.0.93 Patch Release: The Ecosystem Signals Behind a Dependency Update
@ai-sdk/vercel 2.0.93 patch released, syncing @ai-sdk/openai-compatible to 2.0.91. This article analyzes the real impact of this dependency update on the AI SDK ecosystem and developers.
Tech FrontiersVercel AI SDK Vue Package Releases Version 3.0.324: Synchronized Upgrade of ai Dependency
Vercel AI SDK's Vue adapter package @ai-sdk/vue releases patch version 3.0.324, synchronizing the ai core package to 6.0.324. This article covers the dependency sync update, upgrade advice for Vue developers, and the evolution trends behind AI SDK's multi-framework version management strategy.
Tech Frontiers@ai-sdk/xai 3.0.150 Patch Release: Just a Dependency Sync
@ai-sdk/xai 3.0.150 is released as a patch update syncing the @ai-sdk/openai-compatible 2.0.91 dependency, with no API changes — safe for developers to upgrade.