Hacker News: syumei

Multi-Agent Design Patterns – Part 1: Separating Coder and Tester

syumei — Mon, 20 Jul 2026 20:54:51 +0000

Article URL: https://medium.com/@Koukyosyumei/understanding-multi-agent-design-patterns-part-1-agentcoder-separating-code-implementation-from-0e0d2491bec0

Comments URL: https://news.ycombinator.com/item?id=48984759

Points: 3

# Comments: 0

Tile-pack repos, files, and prompts into images for 60–75% fewer Claude tokens

syumei — Fri, 17 Jul 2026 17:13:03 +0000

Article URL: https://old.reddit.com/r/ClaudeCode/comments/1uz5tpl/tilepack_repos_docs_files_and_prompts_into_images/

Comments URL: https://news.ycombinator.com/item?id=48949741

Points: 1

# Comments: 0

I Reimplemented the Workflows of 40 Multi-Agent LLM Papers – Here Are Lessons

syumei — Thu, 16 Jul 2026 02:12:05 +0000

Article URL: https://old.reddit.com/r/vibecoding/comments/1uxpox5/i_reimplemented_the_core_workflows_of_40/

Comments URL: https://news.ycombinator.com/item?id=48929623

Points: 3

# Comments: 0

Show HN: H5i-Python: Python SDK for Programmable Multi-Agent Orchestration

syumei — Wed, 15 Jul 2026 22:25:13 +0000

h5i-python is the Python SDK for the h5i orchestra engine. This SDK lets you define and execute multi-agent coding workflows across Claude Code, Codex, and other runtimes as ordinary Python programs.

Comments URL: https://news.ycombinator.com/item?id=48927933

Points: 2

# Comments: 0

How to Quantitatively Evaluate Prompt Quality in Claude Code and Codex

syumei — Mon, 13 Jul 2026 20:32:23 +0000

Article URL: https://medium.com/@Koukyosyumei/how-to-quantitatively-evaluate-prompt-quality-in-claude-code-and-codex-47f2f27f4bc5

Comments URL: https://news.ycombinator.com/item?id=48898409

Points: 2

# Comments: 0

A Cursor Sandbox Escape Shows Why AI Agents Need Kernel Boundaries

syumei — Tue, 07 Jul 2026 20:24:09 +0000

Article URL: https://medium.com/@Koukyosyumei/a-cursor-sandbox-escape-shows-why-ai-agents-need-kernel-boundaries-281efdb56396

Comments URL: https://news.ycombinator.com/item?id=48823200

Points: 1

# Comments: 0

New comment by syumei in "A Conflict-Free Multi-Agent Ensemble for Claude and Codex"

syumei — Sat, 04 Jul 2026 11:41:19 +0000

Thank you so much!

A Conflict-Free Multi-Agent Ensemble for Claude and Codex

syumei — Sat, 04 Jul 2026 03:18:25 +0000

Article URL: https://medium.com/@Koukyosyumei/a-conflict-free-multi-agent-ensemble-for-claude-and-codex-0ded61c5a35e

Comments URL: https://news.ycombinator.com/item?id=48782358

Points: 1

# Comments: 2

CurXecute: When a Slack message rewrites your editor's config and runs code

syumei — Fri, 03 Jul 2026 15:35:28 +0000

Article URL: https://medium.com/@Koukyosyumei/curxecute-when-a-slack-message-rewrites-your-editors-config-and-runs-code-f7f7907cf0a9

Comments URL: https://news.ycombinator.com/item?id=48776263

Points: 3

# Comments: 0

Claude Code SOCKS5 Bypass Shows Why Egress Filters Belong at the Boundary

syumei — Thu, 02 Jul 2026 15:40:38 +0000

Article URL: https://medium.com/@Koukyosyumei/claude-codes-socks5-proxy-bypass-why-egress-filtering-must-happen-at-the-boundary-aaa445019e69

Comments URL: https://news.ycombinator.com/item?id=48763253

Points: 1

# Comments: 0

Reached 440+ stars: Auditable sandbox to record what AI agents did

syumei — Thu, 02 Jul 2026 02:23:29 +0000

Article URL: https://old.reddit.com/r/SideProject/comments/1ul5fe9/reached_440_stars_built_an_auditable_sandbox_that/

Comments URL: https://news.ycombinator.com/item?id=48755675

Points: 2

# Comments: 0

Auditable Workspaces for AI Coding Agents

syumei — Wed, 01 Jul 2026 13:54:34 +0000

Article URL: https://medium.com/@Koukyosyumei/auditable-workspaces-for-ai-coding-agents-de00eff5f9b9

Comments URL: https://news.ycombinator.com/item?id=48746845

Points: 1

# Comments: 0

I built an anime-style UI for watching AI coding agents review each other's code

syumei — Mon, 29 Jun 2026 14:03:14 +0000

Article URL: https://old.reddit.com/r/codex/comments/1uit86f/i_built_an_animestyle_ui_for_watching_ai_coding/

Comments URL: https://news.ycombinator.com/item?id=48719426

Points: 3

# Comments: 0

New comment by syumei in "Claude Code and Codex Sandbox Untrusted Code"

syumei — Sat, 27 Jun 2026 21:25:57 +0000

Claude Code and Codex both run untrusted commands. This post compares how their sandboxes protect your repo, secrets, and host machine.

How Claude Code and Codex Sandbox Untrusted Code

syumei — Sat, 27 Jun 2026 21:25:57 +0000

Article URL: https://medium.com/@Koukyosyumei/how-claude-code-and-codex-sandbox-untrusted-code-ba39b493046a

Comments URL: https://news.ycombinator.com/item?id=48701924

Points: 2

# Comments: 1

Two Heads Are Better Than One: Run Many AI Agents, Merge One Auditable Result

syumei — Mon, 22 Jun 2026 15:26:23 +0000

Article URL: https://medium.com/@Koukyosyumei/two-heads-are-better-than-one-run-many-coding-agents-merge-one-auditable-result-de58be3a1b1b

Comments URL: https://news.ycombinator.com/item?id=48631543

Points: 2

# Comments: 0

New comment by syumei in "Improving token efficiency in GitHub Copilot"

syumei — Wed, 17 Jun 2026 14:51:21 +0000

quite interesting post

How Claude Code Broke My Git Worktree

syumei — Wed, 17 Jun 2026 14:36:02 +0000

Article URL: https://medium.com/@Koukyosyumei/how-claude-code-completely-broke-my-git-worktree-fc74effc9c4e

Comments URL: https://news.ycombinator.com/item?id=48571146

Points: 3

# Comments: 0

A Git/LFS-backed unified tool-output format for AI agents, with 95% token cut

syumei — Mon, 08 Jun 2026 17:39:16 +0000

Article URL: https://old.reddit.com/r/codex/comments/1u0amqg/a_gitlfsbacked_unified_format_of_tool_output_for/

Comments URL: https://news.ycombinator.com/item?id=48448467

Points: 2

# Comments: 0

New comment by syumei in "Claude Code and Codex can have real-time conversation via Git"

syumei — Thu, 04 Jun 2026 18:08:58 +0000

Fair criticism. I don’t think the value is just "agents can send text to each other”; that part has many implementaton and design choices. The value of h5i I’m exploring is making the intermediate state reviewable: review requests, risks, handoffs, unresolved claims, associated prompts and AI-to-Ai conveersation, and final decisions tied to the branch/PR.