oxfeeefeee

ETH succeeded by offering products that BTC simply cannot. Not erc20 token, but DEFI. Uniswap is the most important thing happened to ETH. SOL has the ability to be the next disruptive L1, but SOL's Uniswap moment is yet to come. Remember, Uniswap was born when ETH was "dead".

@Piglei 你确定没有提高10倍？到目前为止50天不到： volang.dev

40 年前，图灵奖得主 Fred Brooks 说“没有银弹”，指“没有任何一项技术或方法可使软件工程的生产力在十年内提高十倍”。 40 年后，软件开发者已然分为两派。第一派仍相信没有银弹，另一派则坚信银弹已经出现，造成这种分歧的关键原因，是前者读（Review） AI 写的代码，后者则完全不读。

@qinhualai #playground" target="_blank" rel="nofollow noopener">volang.dev/#playground

到底是有多少人全程靠AI做出一款靠谱产品的，敢发出来吗

@discountifu skill issue volang.dev

目前 Vibe Coding 真实有用的场景总结： 1. 编写 tampermonkey 脚本。tampermonkey 是一个权限非常高的插件，尤其对于浏览器里安装了各种钱包的加密货币用户来说，安装来自第三方的脚本危险系数极高，所以我现在所有的工具脚本全都自己写，比如说自动删除google搜索结果广告的、自动把网页转换成markdown打包下载的、自动去除AI生图水印的，这些脚本对于目前的AI模型来说基本都是手到擒来 2. Chrome 插件。对于稍微复杂一些的需求，简单的脚本可能无法满足，但是如果你用 ChatGPT 或者 Manus 一类自带 Agent 模式的工具，它可以直接开发好一个 Chrome 插件并且打包成 zip，你下载解压以后直接就能用了 3. Python 脚本。主要是配合无头浏览器做一些爬虫性质的脚本，定时运行脚本、告警推送，AI 写这类单文件功能也几乎不会犯错，非常好用 4. SEO 优化。直接把博客站点整个丢给 AI，让它自己遍历所有的文章内容页，分析文本打上关键词、标签之类的，亲测稍微有一丁点效果。 虽然我承认目前的 AI 已经很强了，但是在中型及以上的项目中，AI能力的表现还是捉襟见肘，很容易出现把聋子治成哑子的情况，捡了芝麻丢了西瓜，尤其在模块重构的时候，几乎很难一次成功，当然人类程序员在重构屎山代码的时候也需要耗费很多精力，作为一名程序员，我不害怕，反倒是很期待 AI 能够真正彻底取代我的那天早点到来

The first Tetris implemented in Go what runs in a browser locally #playground" target="_blank" rel="nofollow noopener">volang.dev/#playground

@QwQiao @naval The key is, AI gives you infinite "brain horse power" so that you don't need to compromise when writing code, which makes the overall quality better not worse

@QwQiao @naval Even now human can write better machine code than compilers, but that doesn't mean compilers produce bad machine code.

We’re now in the era of slop code.

Agreed

@QwQiao Same，made something better than I was ever able to make github.com/oxfeeefeee/vol…

been coding nonstop during the holidays after years of hiatus. this shit is srsly addictive with the latest ai coding tools and i feel 10x more powerful than i was 3yrs ago.

@hindookissinger true, see what a prompt engineer can do in 20 days :github.com/oxfeeefeee/vol…

To show that vibe coding is better than manual coding. Volang will be the first language to support four backend modes: Hot VM for config and development, Static VM for a small footprint, Static VM-JIT for flexibility, and AOT for native-level performance.

Volang — Golang-ish, Rust-built, VM & JIT, vibe-coded in 20 days github.com/oxfeeefeee/vol…

So the simplest task is writing a parser for a programming language—even though, to humans, this is often considered quite difficult.

Whether AI can write good code depends entirely on whether the requirements are well defined—whether explicitly, through formal specs, or implicitly via shared conventions, aesthetics, or commonly accepted “standard values” of demand.

Just as we once needed humans to work through calculus problems step by step, today you only need to understand calculus—you can use a computer to instantly get answers that used to take a long time to compute. You don’t need to know how to do the actual calculations at all.

With Opus 4.5, we no longer need average programmers—only a few architects. This is so huge that I can’t sleep, and I don’t see people talking about it.

@shanshan521 @kernel1983 目前1998个，solanabeach.io/validators

@kernel1983 排序器只有一个节点，sol还是有好几个节点的哈哈哈

Solana最近异军突起，很多sol持有者们在喊flip eth了。理由是： 比去中心化，sol比eth L2强； 比速度，sol比eth L2高； 比成本，sol比eth L2低。 当你选择用什么区块链的时候，最重要的就是这三点吧？ 我一直觉得以太坊搞L2是在搞不定一层扩容技术的 情况下的权宜之计。换句话说，如果以太坊不能在这个周期内在一层扩容方面实现突破，那前有solana，后有各种btc L2，有点难了。

@mindaoyang @ohyishi - 全链上基于账单本的交易所 @PhoenixTrade - 在自己的虚拟机上再跑一个虚拟机，依然相当快， @Neon_EVM. the list goes on and on

@mindaoyang @ohyishi 在疯狂打铭文的时候，依然能保持其他用户不受干扰，可以以$0.001/tx的费率做交易

有什么是现在 ETH 能做到，SOL 做不到的？