Issue 27: Machine as Man, Man as Machine?

Knowledge is power, info-gap is money!

record (in sports etc)Indie Developers Going OverseasHelping indie developers with quality information that they getNarrowing the information gap, released every Monday.

The weekly magazine is open for submissions and welcomesposting (e.g. to a newsgroup or mailing list)The following:

  • open source project
  • Creative tools/productivity tools
  • Tutorials for Indie Developers Going Overseas
  • Quality articles/tweets/videos

This weekly magazine is published by@weijunextscheduled service (train, bus etc)

Sponsored Listings:

  • Mkdirs: A fully functional nav site template with integrated login, payment, blogging, etc. It took someone as fast as 2 hours to publish their nav site.
  • Next.js documentation: Next.js v15 latest Chinese documentation, website style and official website, creating an immersive Next.js Chinese learning experience.
  • Chrome Plugin Full Stack Development in Action.: Practical teaching class for real offshore projects to help you become a full-stack offshore engineer in half a month.

Machine as man, man as machine?

Yesterday, Li Jigang posted a dynamic in Instant:

I’m going to give a shout out to this dynamic, because I had a similar feeling the other day. I was using an unfamiliar Shadcn component to get Claude to do what I wanted directly on top of a Cursor, and after I wrote it, I realized, if the AI wrote this feature without me knowing anything about it, is it still a feature I’m developing? Obviously not.

If I take credit for developing this feature myself and advertise it on social media as “I made xxx feature in 10 seconds”, am I switching roles with the AI? The AI becomes the brain, it’s the one who thinks, and I become the ‘tool’ doing the repetitive work – hitting the tab key again and again and posting to social media.

Those keen to claim the fruits of AI production as their own will find themselves falling into the trap of instant gratification:

  • Surface understanding: over-reliance on AI to generate code leads to a superficial understanding of the problem
  • Degradation of capacity: problem-solving thinking muscles are gradually atrophied by lack of exercise
  • False confidence: AI-generated code gives us the false impression that we’ve mastered a technology

In fact, after using AI programming in depth, it’s easy to find that AI can perfectly realize a single function, but it can’t do anything about complex needs, and it needs the human brain to plan and split. For example, recently I used AI to do automatic translation and verification scripts, once the script exceeded 500 lines, it was like a mountain of shit, I couldn’t change it, and neither could AI.

“AI has limitations” is the consensus on the Internet, which means that humans need to maintain their professionalism in order to truly command AI. if even the basics are lost, and we are satisfied with instant feedback, the result is likely to turn out to be an exchange of roles between us and AI.

Therefore, I am also writing a word of encouragement with you in the immediate future:

open source project

  • PromptFix

    A very effective AI image repair tool that supports a variety of repair functions:

    • Image coloring
    • low-light enhancement
    • Watermark Removal
    • Object removal
    • Snow removal
    • 去雾
    • Deblurring Pre-trained model weights can be downloaded from HuggingFace, and the open source repository provides the full training and inference code.

    promptfix

  • JSON i18n Translation Tool

    A JSON translation tool that supports 40+ languages, the author open-sourced a complete AI-assisted PRD documentation + source code based on Next.js, for those who do not write code but want to develop the product has great reference value.

    JSON

  • Better Auth

    A new authorization library is out. Better Auth is a framework-agnostic TypeScript authentication (and authorization) library. It provides a complete set of functionality out-of-the-box and includes an ecosystem of plugins that allow you to add advanced functionality in a short period of time with minimal code.

    better-auth

  • Markdown To Poster

    This React component is used to render Markdown into beautiful social media images. In addition, the project includes a built-in WEB Editor that can be deployed with a single click and used as an online Markdown to poster image editor.

    markdown-to-image

  • MinerU

    Born out of InternLM’s pre-training process, MinerU is a high-quality PDF conversion tool that can:

    1. Preserve the complete structure of the original PDF document (including headings, paragraphs, lists, etc.)
    2. Intelligent extraction of various types of content (including text, images and their explanatory text, tables and their captions, and mathematical formulas)
    3. Automatic format conversion
    4. Built-in intelligent OCR function
    5. Supports multiple output formats such as Markdown, JSON, etc.
  • Thinking-Claude

    By adding a thought chain to Claude as well, the answer will show the reasoning process as well, and the accuracy becomes much better.

    Usage:

    1. Copy the latest version of the reflection protocol from the model_instructions folder
    2. Creating a new project in Claude.ai
    3. Paste the command into the Custom Commands section
    4. Install browser extensions (Chrome supported, Firefox version in development)

Tool Recommendations

  • GitHub Receipt

    This is interesting, generate a receipt with your GitHub info.

    github-receipt

  • onlook

    Lets you visualize and design your React UI like Figma, runs natively, and integrates with Tailwind CSS.

    onlook

  • Git Cheat Sheet

    git-cheat-sheet

Going to Sea Tutorials/Technical Tutorials/Resource Sharing