@howie_serious: https://x.com/howie_serious/status/2054778826006347949
Summary
The author shares an in-depth experience using Codex Agent, including the macify plugin update, AI Insider 2.0, and the wechat2readwise service, emphasizing its powerful features and low cost.
View Cached Full Text
Cached at: 05/14/26, 06:31 AM
codex agent: So Powerful It’s Almost Dangerous, So Cheap It’s Almost Free
In the past few days, I’ve been completely absorbed in codex.
Daily token consumption ranges from at least 100 million to nearly 300 million.
In less than 7 days, I’ve used 1.3 billion tokens, which would cost about $1,100 if billed via API.
Of course, OpenAI built its trillion-dollar AI factory where token production costs are close to electricity bills, so they heavily subsidize it. A $200/month Pro subscription effectively lets you spend $5,000 worth of tokens per month.
What did I build?
macify Plugin
-
Added two new features to Achan’s Chrome extension macify. My first time merging code on GitHub!
-
Built a third-party website for macify: macify.candobear.com
AI Insider 2.0
- Upgraded AI Insider to version 2.0: launched the site seriousai.candobear.com, which can be added as a PWA app on both mobile and desktop.
Added semantic search capabilities, seamlessly integrating AI into the AI Insider site. For example, searching for “财富密码” (wealth secrets):
Also rebuilt the entire insider workflow inside codex: faster, better, more powerful — and nearly free.
- Insider MCP
AI Insider is now fully integrated into agents and chatbots, with new additions: Insider CLI, Insider MCP, Insider SKILL, and Insider AGENT.
You can use it in any chatbot or agent! And explore new features like topic research and cross-article reading.
wechat2readwise
- wechat2readwise service: Yesterday I tackled a high-difficulty project — building a readwise inbox service within the WeChat ecosystem.
It involved WeChat service account development, WeCom (WeChat Work), WeChat Pay, Tencent Cloud, Cloudflare… but everything went smoothly, going from zero to a satisfying iteration in under 24 hours.
Try it: wechat2readwise.candobear.com
codex: So Powerful It’s Almost Dangerous
ChatGPT and the LLMs it represents are often called a “GPT” technology: General Purpose Technology. As the name suggests, master this technology and you can solve any problem.
The revolutionary new version of codex, updated on April 16, 2026, truly embodies the meaning of “GPT”: for (almost) everyone, for (almost) everything!
So powerful it’s almost dangerous.
So cheap it’s almost free.
That’s why I’ll spend 2 hours in a codex special session, sharing heavy practical experience — to talk with you about the codex agent!
Lecture Outline
Content is rich, but the structure follows the principle of Dao, Law, Technique, Tool, combined with real-world cases and hands-on practice.
Livestream Room
Scan the QR code to enter the livestream room.
Similar Articles
@gengdaJ: https://x.com/gengdaJ/status/2053724702993190917
This tutorial details the advanced uses of Codex App, including generating office documents, creating 3D videos, deploying websites, and other practical scenarios to help users enhance work efficiency and automation.
@justloveabit: With This Open-Source Tool, I Got a Team of AIs to Work for Me. Here's the deal: I've been tinkering with various AI agents lately. Multiple Claude Code windows open, Codex running, occasionally using Cursor. The result? Total chaos—I had no idea what each agent was doing or how much it was costing. Restar…
This article introduces Paperclip, an open-source tool designed to centrally manage and orchestrate multiple AI agents. By simulating a corporate organizational structure, task assignment, and budget control, it addresses key pain points in multi-agent collaboration, such as lost context, unpredictable costs, and chaotic scheduling.
@Lonely__MH: https://x.com/Lonely__MH/status/2055156505796866407
This article shares seven tips for configuring the Hermes AI agent, including setting its soul, memory system, skill solidification, scheduled tasks, auxiliary models, multi-platform gateways, and mobile phone control, and introduces how to use models provided by Nous Research for free.
@berryxia: Codex has now transformed itself into a factory capable of building AI products autonomously. OpenAI officially announced today: Codex now includes an 'OpenAI Developers' plugin, allowing you to create API keys and call all OpenAI APIs directly from the terminal. Automatically turning ideas into fully functional...
OpenAI announced that Codex now features an 'OpenAI Developers' plugin, supporting the creation of API keys and invocation of all APIs directly within the terminal, enabling automated construction from idea to runnable AI applications and agents.
@VincentLogic: Found a pretty interesting AI assistant client! Hermes Agent, a clean-looking Chinese desktop app. Feature integration is quite comprehensive: - Conversation & Session Management - Multi-model Support - Skill & Tool Integration - Scheduled Tasks & Gateway Configuration From the interface, it can help you: Search web, set reminders, summarize emails,…
Hermes Agent is a cross-platform AI assistant desktop client developed based on Electron. It supports multi-model switching, skill integration, scheduled tasks, and more, aiming to provide users with a unified AI productivity workspace.