I feel like I previously fell into an AI programming dead end.
Over the past while, I used Codex to build a lot of things:
Meme trading tools, contract strategies, arbitrage monitoring, and all kinds of automation scripts... It looks like I’m producing a lot, and the code keeps growing, but when I look back, the things that can reliably generate value aren’t really there—I’ve only managed to get back my AI subscription fee.
After reorganizing things again these past few days, I realized there are already many mature open-source projects on GitHub. Trading frameworks, on-chain data, wallet analysis, arbitrage monitoring, backtesting, execution engines, all kinds of SDKs and APIs... A lot of these have been iterated on by others for years, even validated in real environments. There’s no reason for you to have the AI “recreate the wheel.”
Before, building a system was expensive, so people would seriously consider whether “it’s worth developing.” Now AI lowers the cost of writing code too much, and it’s easy to develop a delusion: anything you can think of suddenly feels worth building.
Code cost may be lower, but time cost, validation cost, maintenance cost, and trial-and-error cost haven’t disappeared. In fact, the faster AI writes code, the easier it is for a wrong direction to rapidly balloon into tens of thousands of lines of code.
The enthusiasm for “vibe coding” fades, and reality is a mess all over the place
#中国八大金融机构注资3600亿元
$BTC $BNB
Over the past while, I used Codex to build a lot of things:
Meme trading tools, contract strategies, arbitrage monitoring, and all kinds of automation scripts... It looks like I’m producing a lot, and the code keeps growing, but when I look back, the things that can reliably generate value aren’t really there—I’ve only managed to get back my AI subscription fee.
After reorganizing things again these past few days, I realized there are already many mature open-source projects on GitHub. Trading frameworks, on-chain data, wallet analysis, arbitrage monitoring, backtesting, execution engines, all kinds of SDKs and APIs... A lot of these have been iterated on by others for years, even validated in real environments. There’s no reason for you to have the AI “recreate the wheel.”
Before, building a system was expensive, so people would seriously consider whether “it’s worth developing.” Now AI lowers the cost of writing code too much, and it’s easy to develop a delusion: anything you can think of suddenly feels worth building.
Code cost may be lower, but time cost, validation cost, maintenance cost, and trial-and-error cost haven’t disappeared. In fact, the faster AI writes code, the easier it is for a wrong direction to rapidly balloon into tens of thousands of lines of code.
The enthusiasm for “vibe coding” fades, and reality is a mess all over the place
#中国八大金融机构注资3600亿元
$BTC $BNB
