Stop guessing whether your Skill works: skill-optimizer measures and improves it

Explore all

Three Kinds of AI Agent. Only One of Them Lives in Your Browser.

Explore three types of AI agents, focusing on browser-native agents and their potential to transform AI workflows by operating directly within web environments.

WebMCP: Making Web Apps Faster and Cheaper for AI Agents

WebMCP is a proposed API standard by Chrome to improve AI agent interactions with web apps by providing interfaces that communicate capabilities, reducing errors and costs.

With Cosmos, Augment Code wants to give AI coding teams shared memory and context

Cosmos by Augment Code provides AI coding teams with shared memory and context, enabling coordinated agent tasks and knowledge reuse across development environments.

GPT-5.5 is OpenAI's best model. But paying more for it makes no sense.

GPT-5.5 is OpenAI's most capable model, but its 63% higher cost offers minimal performance gains over GPT-5.4, making it less cost-effective for most tasks.

From Blind Spots to Merged PRs: Why Your Coding Agents Need Runtime Intelligence

Explore why coding agents need runtime intelligence to avoid costly production errors, focusing on the gap between static code understanding and real-world execution.

Warp goes open source, betting agents and community can outpace closed rivals

Warp open-sources its terminal client, leveraging AI agents and community input to redefine software development, focusing on task coordination and output verification.

OpenAI open-sources Symphony, a spec for orchestrating Codex agents

OpenAI's Symphony, an open-source spec, orchestrates Codex agents for coding tasks, automating workflows from issue tracking to pull requests with human review.

Common Pitfalls of Skills Development (And How to Fix Them)

Explore common pitfalls in AI skills development, including outdated agent config files, and learn strategies to address these issues for better procedural knowledge management.

Anthropic postmortem shows how small changes compounded into Claude Code failure

Anthropic's postmortem reveals Claude Code's decline was due to prompt changes, caching bugs, and reasoning adjustments, highlighting how small system tweaks can impact AI tools.

As SpaceX deal looms, Cursor partners with Chainguard to secure open-source dependencies in AI-built code

Cursor partners with Chainguard to secure open-source dependencies in AI-generated code, addressing supply chain risks as AI tools automate software production.

Replit launches “Security Agent” to scan and fix vulnerabilities in AI-built apps

Replit's new 'Security Agent' integrates automated vulnerability detection and remediation into its AI coding platform, enhancing security during app development.

Evaluating Kimi 2.5 vs Kimi 2.6: What happens to agent skills when the model gets smarter?

Early signals from benchmarking Kimi K2.5, K2.6, and Sonnet 4.5 on 21 agent skills. Kimi K2.6 is a better model than K2.5, and skills still matter as models improve.