Vercel eve Makes the Filesystem the Interface for Durable AI Agents
Vercel's open-source eve framework turns agents into inspectable directories, then supplies durable execution, sandboxes, approvals, channels, subagents, and schedules.
6 min read
Topics9 articles
Open Source
Focused ASHFOG writing connected by this subject.
Cloudflare Open-Sourced the Security Audit Skill Behind Its Vulnerability Harness
Cloudflare's security-audit-skill turns coding agents into a structured audit pipeline with coverage tracking, adversarial validation, and machine-readable findings.
5 min read
YOINK Turns an Obsidian Vault Into a Second Brain That Grades Your Predictions
YOINK keeps notes as plain Markdown, adds falsifiable claims with confidence scores, and uses scoring to show whether your second brain is improving your judgment.
5 min read
Mojo 1.0 Is Here: Is It Finally Ready for the AI Programming Era?
Mojo 1.0 stabilizes the language, opens the compiler, and targets heterogeneous AI hardware—but the real question is where it already makes sense to use it.
6 min read
ThreeUI Is Open Source: Build 3D Websites Without Starting From Scratch in Three.js
ThreeUI Community turns reusable Three.js scenes, shaders, hero sections, and interactive effects into copy-ready building blocks with an MIT-licensed open-source catalog.
5 min read
OpenOPC Treats AI Agents Like Employees—and the Company Is the Runtime
OpenOPC is an open-source attempt to put organization, delegation, review, memory, and approval above interchangeable AI agents such as Codex and Claude Code.
7 min read
AnyDoc Turns Office Files Into Fast, Local Markdown for AI Pipelines
Firecrawl’s Rust-based AnyDoc converts fourteen office and document formats into consistent Markdown, creating a fast local ingestion layer for RAG and AI agents.
10 min read
Five Agent Infrastructure Projects That Solve What Bigger Context Windows Cannot
Beads, Crush, Claude Context, TradingAgents and ML Intern each address a different failure mode in AI work: memory, interface, retrieval, coordination and execution.
10 min read
OpenWorker Makes the Personal Desktop Agent a Local, Model-Agnostic Tool
Andrew Ng’s OpenWorker is an MIT-licensed desktop coworker that turns outcome requests into files, messages and scheduled work while keeping models and permissions under user control.
9 min read