The picks and shovels of the AI gold rush: SDKs, evals, observability, frameworks and infra that builders actually use.
331 articles
Vercel Pro and Enterprise teams can now select Basic build machines with 2 vCPUs and 8 GB of memory. They provide a lower-resource, cost-efficient option for smaller applications and AI agents, while Elastic build machines remain the default for new projects.
Vercel News · 17d ago
Hugging Face presents an approach for giving coding agents persistent memory that remains under the user's control. The update focuses on improving continuity and personalization across coding-agent sessions.
Hugging Face Blog · 17d ago
GitHub explains why shorter model outputs can sometimes cost more across an entire coding task. The post describes how GitHub Copilot reduces wasted work while preserving task quality.
GitHub Blog · 17d ago
Apple now allows Mac App Store apps that require macOS 13 or later and use the universal format to drop support for Intel-based Macs. Developers can simplify their build process and reduce download and on-device storage requirements.
cnBeta (Full RSS) · 17d ago
Developers are repurposing Apple’s discontinued Touch Bar as a compact status panel for long-running AI coding and research agents. Tools such as T1Bridge, Agent Status for Pock, and Codex Touch Bar Buddy show task states and let users jump back to active workspaces without monitoring the main screen.
虎嗅 · 18d ago
Google is testing an official Flatpak packaging path for Linux versions of Chromium and Chrome, alongside existing RPM and DEB packages. The effort must resolve compatibility issues between Flatpak's sandbox and Chromium's built-in browser sandbox, with no confirmed release date yet.
IT之家 · 23d ago
OpenAI will reset Codex usage limits again after confirming that some users are consuming their allotted usage faster than expected. The change is intended to address unexpectedly rapid usage depletion.
ITmedia AI+ (日本) · 27d ago
JetBrains has outlined its initial measures for managing rapidly increasing AI spending. The update signals a shift toward tighter cost control as AI becomes more deeply integrated into JetBrains products and operations.
InfoQ中国 · 29d ago
Qwen Code’s nightly release adds messaging that explains why a review loop has not settled. It also fixes a CI fallback comment that could incorrectly deny a review it had already posted.
Qwen (GitHub Releases: qwen-code) · 30d ago
Vercel Functions can now emit custom metrics and analyze them alongside built-in observability data. Developers can filter, aggregate, visualize, and query these metrics across Vercel Observability.
Vercel News · 30d ago
The article argues that a Coding Agent’s practical value depends heavily on how well it understands a developer’s codebase, workflows, and organizational data. It presents data access and context quality as central factors in improving coding-agent performance.
InfoQ中国 · 30d ago
This analysis examines why Anthropic has gained an early compounding advantage from its focus on coding, despite many companies pursuing the same direction. It focuses on how model capabilities and product design may evolve together to strengthen that advantage.
钛媒体 · 30d ago
Cognition's CEO denied a report claiming that SpaceX was in talks to acquire the AI coding startup. The report comes as SpaceX reportedly expands its enterprise AI ambitions and has already acquired Cursor.
TechCrunch AI · 31d ago
Algolia is now available for provisioning and management through the Vercel Marketplace. The integration supports hosted search, AI ranking, recommendations, crawler-based indexing, and automatic environment-variable configuration.
Vercel News · 31d ago
The article highlights how users are turning DeepSeek Harness into a modular ecosystem by combining community-built plugins. It curates 11 notable plugins, framing the tool as a flexible platform for recreating different workflows and applications.
Ifanr (爱范儿) · 33d ago
The article examines MCP’s move toward a stateless architecture and the questions it raises among developers. The central debate is whether reducing state makes MCP more scalable or simply brings it back to the conventional API model.
InfoQ中国 · 34d ago
The article presents ASCF, the development framework behind HarmonyOS atomic services, as a way to reduce development costs and improve delivery efficiency. It argues that the framework can connect faster implementation with stronger commercial returns for developers.
量子位 · 35d ago
Qwen Code’s v0.21.11 nightly release introduces a deny-by-default footprint gate and positional window censuses for autofix workflows. It also enables web-shell split-view panes to share @-mention providers.
Qwen (GitHub Releases: qwen-code) · 36d ago
The article examines a proposed DeepSeek + Pi combination that may outperform Claude Code in coding workflows. Pi’s founder says he anticipated the potential of this pairing early.
InfoQ中国 · 36d ago
Qwen Code reports a full end-to-end validation cycle spanning release creation, Actions, DSW SWE-bench Verified 500, publishing, and Terminal-Bench 2.0. The run recorded a Terminal-Bench 2.0 score of 89 and references Benchmark-Qwen-Ref v0.21.2.
Qwen (GitHub Releases: qwen-code) · 36d ago
Bloom Security’s Extension Resurrection research argues that IDE extensions deserve the same scrutiny as software dependencies and build pipelines. The study examined extension packs on the Visual Studio Code Marketplace and Open VSX.
The Next Web (TNW) · 37d ago
A survey of 138 developers found that 75% use Claude Code in their daily coding workflows. The article examines the factors developers say influence their preference over Codex.
ZDNet AI · 37d ago
AI coding agents are lowering the barrier for ordinary users to build highly personalized software, prompting one writer to replace paid Mac applications with custom tools. The article argues that AI accessibility, open file formats, and human product judgment are becoming more important than feature-heavy commercial software.
虎嗅 · 37d ago
The article previews Musk’s return with an AI Agent designed to work continuously around the clock. It attributes the product’s development to members of the Cursor team, but provides no further product specifications.
Ifanr (爱范儿) · 38d ago
GitHub’s tutorial explains how to write a first prompt in the GitHub Copilot app. It covers selecting the right context and model before starting a task.
GitHub Blog · 38d ago
The article argues that AI video is moving beyond standalone generation models toward harnesses that coordinate and operationalize video-model workflows. It positions LibTV as a Codex-like layer for video, following the emergence of Seedance 2.5.
Ifanr (爱范儿) · 40d ago
LangChain compares Deep Agents, LangChain, and LangGraph as distinct open-source approaches for building AI agents. The post explains their key differences and offers guidance on when developers should use each framework.
LangChain Blog · 44d ago
The article claims that AI has challenged CUDA’s long-standing software advantage by reproducing or bypassing part of its ecosystem in roughly 10 hours. It raises the question of whether NVIDIA’s CUDA moat is becoming more vulnerable, but provides no technical details about the specific method.
量子位 · 45d ago
Alibaba is preparing for a major milestone in its AI coding development, aiming to maintain its competitive edge. The company is currently in a critical sprint phase to secure its position as a leader in the AI coding space.
钛媒体 · 50d ago
Vercel has introduced a minimap feature to its Workflows trace viewer, allowing users to visualize entire workflow runs at a glance. The tool supports intuitive navigation, including panning, zooming, and jumping to specific execution sections.
Vercel News · 67d ago