大模型和编码代理能力的提升正在改变软件开发的可行性边界。1随着AI工具的进步，原本需要数年才能完成的项目现在可在数小时内实现。1作者通过多个实例证明了这一转变：用数小时而非数年开发出正则表达式引擎、ripgrep分支版本和Rust编译加速工具。1值得注意的是，pgrust项目在测试中发现了Postgres中的多个真实bug，包括数值bug、格式缓存bug和pgcrypto bug，表明AI辅助开发可以产生具有实际价值的成果。1
AI工具的能力已经达到了令人瞩目的水平。1使用GPT-5.1/5.2约5小时就能开发出超人类水平的棋类AI，而后续模型优化进一步加快了这一过程。1在zstd实现测试中，GPT-6 Astra的成功率超过99%，仅需约1美元的成本。1Anthropic的七人团队甚至在五个月内使用AI构建了GitHub替代系统及多个相关工具。1
尽管当前许多AI生成的项目质量参差不齐，但对于拥有专业能力的开发者而言，不应被此阻碍。1开发者应当构建更具雄心的项目，而非盲目追随潮流，充分利用现代AI工具所提供的机遇。1
The capabilities of large language models and coding agents have fundamentally changed what is feasible for software developers to accomplish, according to a discussion on Hacker News.1 Modern AI tools now compress work that previously required years into mere hours, enabling developers with professional expertise to pursue projects that were once impractical.1
Several concrete examples illustrate this shift. A project called pgrust discovered multiple genuine bugs in PostgreSQL during testing, including numerical bugs, format cache bugs, and pgcrypto bugs.1 Using GPT-5.1 and 5.2 models, a developer created a superhuman-level chess AI in approximately five hours—work that would have taken far longer through traditional methods—and later models further accelerated this process.1 GPT-6 Astra achieved a success rate exceeding 99 percent when tested on zstd implementations, requiring only about $1 in computational costs.1 Additionally, a seven-person team at Anthropic built a GitHub alternative system along with multiple related tools using AI assistance over a five-month period.1 Other projects completed in hours rather than years include custom regular expression engines, ripgrep forks, and Rust compilation optimizations.1
While acknowledging that many AI-generated projects currently suffer from poor quality, the author contends that capable developers should not follow prevailing trends.1 Instead, they should focus on constructing genuinely ambitious projects that leverage these powerful new tools.1
评论
还没有评论，欢迎留下第一条。