15:I[981,["658","static/chunks/658-a698693f56d0116a.js","183","static/chunks/183-6beeeb801c274aaa.js","474","static/chunks/app/%5Btype%5D/layout-e8e83705b5c8d3b7.js"],"Posts"] 16:T6ca,Hi HN, I’m building AgentWing.

The idea is simple: AI agents usually execute work as one long sequence:

plan → act → wait → observe → act again → verify → finish

That makes agent workflows feel slow, especially when parts of the task do not actually depend on each other.

AgentWing adds a director layer that turns one agent task into multiple isolated worker runs. The director decides which parts can run at the same time, assigns each worker its own scoped job, runs them in parallel, then verifies and combines the result.

For an enterprise-grade setup, each worker can run inside its own isolated VM or sandbox, so every agent has its own environment, files, tools, and execution boundary.

Example:

Task: “Research a market and prepare a competitor summary.”

A normal agent may do this sequentially:

- find competitors - analyze pricing - compare positioning - look for user complaints - write summary

AgentWing can run the independent parts in parallel:

- worker A finds competitors - worker B analyzes pricing - worker C checks positioning - worker D gathers user complaints - verifier combines the final result

The goal is to make agent tasks feel much closer to instant by removing unnecessary sequential waiting.

I’m experimenting with two modes:

- split mode: different workers handle different parts of one task - race mode: duplicate workers try the same subtask and the best verified result wins

This is still early. The hard part is the director: deciding what can safely run in parallel, what must stay sequential, and how to verify the final output.

I’d love feedback on where this breaks, what workflows would benefit most, and what demo would best show the speedup.

Yet Another AI Teammate

Intelligent eyewear is coming this fall

Build managed agents with the Gemini API

Lighthouse Attention

Anthropic Announces Managed Agents – what this means for consumers

Why raw blockchain logs are no longer enough

State of Crypto 2025: The year crypto went mainstream

Trump Mobile has a real phone, but it may also have a real data leak

Trusted Publishing for NPM Packages

Show HN: Reckoner – a semantic browser for structured data

Ask HN: Natural Language App Testing?

Google's SynthID AI watermarking tech is being adopted by OpenAI, Nvidia

Agentic PCB Design Sucks

Remove AI Watermarks

Show HN: Remindy – A Spotlight-style reminder app for Mac

Language Models Can Autonomously Hack and Self-Replicate

Censorship Request from Coin Insider (2024)

RemotePower – Self-hosted Linux fleet manager (no database, pure Python)

Show HN: Trophy UI, Open-source gamification components built on Shadcn

Show HN: AgentWing – make AI agents complete tasks faster

Google Antigravity Python SDK

Detecting Relaxed Memory Concurrency Bugs in C and C++ Compilers

IRS Must Drop Audits of Trump and Family

The downgrading semantics of memory safety (Extended version)

Self-Improving Reward Models

AI and the Practical Scientist

React Performance Isn't About UseMemo – It's About Render Boundaries

AgentPTY: Script Claude Code’s interactive CLI through a persistent PTY

Agent First Language by Vercel

Valkey 9.1 trims memory 10% and pulls search into the core

Show HN: Git-chain – Simplify branch dependency management in Git

Comments