Ah Oh

AI Builder

🤖Codey - -

[Ah Oh @itsAhOh]

@ZacharyZhang 太厉害了，LibreOffice 升级版

[Ah Oh @itsAhOh]

I prefer games over PowerPoint slides

[Sac @Saccc_c]

花费16天，我用 Opus 5.5 将小王子的故事制作成了3D剧情游戏

结合Tripo+Blender，Opus 5.5 将小王子故事中的每一个星球变成了可探索的3D场景

希望大家能支持一手，后续会开源整个游戏，以及详细制作流程

Welcome to The Little Prince: Across the Stars✨：https://t.co/qr4wR0jLu3

[Ah Oh @itsAhOh]

@ZacharyZhang 同感，放假的时候会有一种必须要干点什么的感觉。感觉运动和睡觉是唯二能放松的事情。

[Ah Oh @itsAhOh]

Every social media platform should do this. @Reddit

[Steve Derico @stevederico]

the @Omarchy handle now has a built-in Issue Tracker

just tag @Omarchy and @Grok opens a GitHub issue

[Ah Oh @itsAhOh]

My favourite bot team in Codey or any other agents is a Dev Team: a Product Manager, a Developer, and a Reviewer.

A product manager figures out what I am trying to do, a developer turns ideas into code, and a reviewer sees improvement.

I think it makes sense in the human tech teams.

[Ah Oh @itsAhOh]

🤯🤯

[Wilf Lin @Wilf_Lin]

我正在做一些东西，使你可以在Hyprland，Windows和MacOS之间跨系统拖拽窗口

总之还要一些时间，5.6 sol设计了过于严格的安全措施，astra正在清理它们，距离发布应该不会太久了

我不会为这个项目设计任何安装器，你需要让你的Agent安装它。

他将默认支持Arch/Omarchy Hyprland v0.56.2，Windows 11 26H1，MacOS 27

[Ah Oh @itsAhOh]

@ZacharyZhang 很厉害，加油加油！

[Ah Oh @itsAhOh]

The idea of my voice control keyboard is simple.

A voice model to understand what you say and summarize it to actions. Then passing results to a classification mode. Execute.

I use Laya because it is light and open source.

It is not to replace keyboards. I use it to speed up my workflow by reducing searches.

[Ah Oh @itsAhOh]

New Demo: Voice control keyboard

I love keyboard shortcuts. I just don’t love memorizing a different set for every app.

So I’m building a keyboard alternative: say what you want to do, and it triggers the action or shortcut for you. No more playing piano on your keyboard just to switch tabs or move between panes.

The goal is simple: keep using the tools you love, with less muscle memory to maintain. Just say it.

I use a voice dictation voice with a local classification model(Laya, not Jev) to map natural phrases to actions, so you won’t have to remember exact commands either.

It’ll be open source. Hoping to share it in the next few days.

[Ah Oh @itsAhOh]

I love iterm2 and @herdrdev for TUI. Let me know if you have other preferences. I will add them in the first version.

[Ah Oh @itsAhOh]

New Demo: Voice control keyboard

I love keyboard shortcuts. I just don’t love memorizing a different set for every app.

So I’m building a keyboard alternative: say what you want to do, and it triggers the action or shortcut for you. No more playing piano on your keyboard just to switch tabs or move between panes.

The goal is simple: keep using the tools you love, with less muscle memory to maintain. Just say it.

I use a voice dictation voice with a local classification model(Laya, not Jev) to map natural phrases to actions, so you won’t have to remember exact commands either.

It’ll be open source. Hoping to share it in the next few days.

[Ah Oh @itsAhOh]

@yifanxu_ephai 我感觉终究大家在使用上还是会忍不住越加越多。倒是像dsh的plugin的概念可能更好，大家需要什么就组装什么，不需要也可以清理干净。

[Ah Oh @itsAhOh]

Tired of repeating yourself to every coding agent?

Co-memo = one local memory for Claude Code, Codex, OpenCode & Pi.

Tell one agent "this repo uses pnpm" → they all know.
Forget it once → gone everywhere.

No account. No API key. Just SQLite + Markdown.

https://t.co/BNhadsaOHQ

[Ah Oh @itsAhOh]

@ThoughtLinkCo Memories are summarized by agents and I haven’t seen misrepresentation yet. Based on how agents work you might need to remind them of updates. Now that you mentioned it, I think it could be a good feature to add.

[Ah Oh @itsAhOh]

I spent a month fixing how Codey remembers things.

"My AI forgot something" is annoying. "My AI quietly erased what I taught it" is worse. We had that second bug.

Here's what changed:

1. Edits that vanished
memory.md looked like the source of truth. It wasn't. It was rebuilt from index.json every time a new memory was saved, so anything you typed there got wiped. Now you edit the memory entries themselves: add, edit, forget. No more lost notes.

2. One brain for every agent
Claude Code reads CLAUDE.md. Codex, OpenCode and Pi read AGENTS.md. Codey now keeps one shared file and mirrors it into each one, inside a marked block. Your own text around it is never touched. It's off until you turn it on.

3. Switching agents used to crash
Coming back to an agent replayed every message it missed into its prompt. Long gap → prompt too big → the agent never even started. Now short gaps go inline, and long gaps get a pointer to a transcript file the agent can read.

4. Global memory got a home
It used to exist only through /remember --global, so it sat empty. Now it has its own Settings ▸ Memory tab, plus switches to turn memory off.

Lesson: good memory isn't only storing things. It's being able to trust what's stored.

#AIagents #ClaudeCode #DevTools #BuildInPublic

[Ah Oh @itsAhOh]

A few weeks ago, I had a feeling that bot chat could be a trend. Love to see what improvements Manus can bring to the table.

If you want to talk to your local agents like this. Try Codey😄

[Cue Agents @CueAgents]

Hello world!

[Ah Oh @itsAhOh]

@yifanxu_ephai 以前重造轮子很傻，现在从头做一个顺心的产品成本太低了。

[Ah Oh @itsAhOh]

Codey 1.13.0 is out

- Refactored cross-agent memory management. https://t.co/5SMdFTuiq1
- New chat mode: Bot conversation.

[Ah Oh @itsAhOh]

@ThoughtLinkCo I don’t know how other people manage. I would ask agents to remember things, and check what they actually memorized.

[Ah Oh @itsAhOh]

No restart needed once the integration is loaded and hooks are enabled. Claude Code and Codex refresh on the next prompt, Pi before each agent run, and OpenCode on the next model request.

It’s refresh at those boundaries, not a live push into an ongoing response. An agent can also explicitly call memory_context to fetch fresh notes. With hooks disabled, that explicit fetch is required.

[Ah Oh @itsAhOh]

My latest opensource project: Co-memo -- one local memory store for Claude Code, Codex, Pi, and OpenCode.

Why? I kept explaining the same preferences and project decisions to different coding agents.

Personal preferences follow you across projects. Project decisions stay with the right project.

Built with SQLite + editable Markdown.

Save once, recall from another agent. Updates and deletions sync too.

https://t.co/clMfjd65Ta

[Ah Oh @itsAhOh]

One SQLite store, one notes table, with explicit scope and project_id fields—not path prefixes. Personal preferences use user; project decisions carry a project ID.

Each agent gets its own Markdown replica. Tool writes are serialized and version-checked; divergent replica edits become explicit conflicts rather than silent overwrites.

Two agents directly overwriting the same Markdown file can still race. The concurrency protection applies to coordinated tool writes, not arbitrary filesystem edits.