Braintrust 是用于交付高质量 AI 产品的可观测性与评估平台。
借助 Codex，其工程师现在可以接收客户功能请求，并在数分钟内创建预览分支，向客户展示可行的创意。
一个月内，Braintrust 团队半数成员已转向使用 Codex。对于创始人兼首席执行官 Ankur Goyal 而言，最大的变化不仅是编码速度的提升，更是与客户之间反馈循环的加速。
“听起来很简单，但 Codex 确实能在终端中输出更多文本而不变慢，其他模型无法复制这一点，”Goyal 表示。
“最大的收益在于速度。”
——Ankur Goyal，创始人兼首席执行官
数分钟内将客户请求转化为预览
速度常被视为工具的一种属性，独立于其核心功能存在，但对 Goyal 而言，速度差异“改变了我与 Codex 互动的方式，与其他模型截然不同。”
借助 Codex，Braintrust 团队能将迭代融入开发工作流，而非让请求搁置等待。“Codex 让我们能够实时尝试客户功能请求，”Goyal 表示，“以前，如果有人提出功能请求，它会进入待办列表，稍后再排优先级。”
如今，团队可以将请求复制粘贴到 Codex，创建预览分支，并在数分钟内将完成的请求展示给客户。“Codex 真正酷的地方在于，我们能实时与客户一起迭代和构思功能请求，”Goyal 说道。
“我们编写的代码越多，能解决的客户问题就越多，而 Codex 是目前实现这一目标最有效的方式。”
——Ankur Goyal，创始人兼首席执行官
速度让自主解决问题成为可能
对 Goyal 而言，Codex 改变了尝试新想法所需的准备工作。“使用其他模型时，我必须尝试提示模型来解决特定问题，”他说。较慢的工具需要更多手动指导，从而提高了实验成本。
“使用 Codex 后，我转而编写一个能演示问题的测试，创建沙盒环境，然后让 Codex 在该环境中运行，”Goyal 表示，“这对我来说是一种新颖的用例，而正是由于速度，我才能进行实验。”
这种速度为 Braintrust 团队提供了更多实验空间。工程师无需逐步提示，而是定义问题，让 Codex 在受控环境中工作，从而更快地从创意走向可行解决方案。
“Codex 真正酷的地方在于，我们能实时与客户一起迭代和构思功能请求。”
——Ankur Goyal，创始人兼首席执行官
Braintrust is the observability and eval platform for shipping quality AI products.
With Codex, its engineers can now take customer feature requests and create preview branches to show working ideas to customers in minutes.
In one month, half of the Braintrust team moved to Codex. For founder and CEO Ankur Goyal, the biggest change is not just faster coding. It’s a faster feedback loop with customers.
“It sounds simple, but Codex can literally print more text in the terminal without getting slow, and other models just can’t replicate that,” says Goyal.
“The biggest gain is speed.”
—Ankur Goyal, Founder and CEO
Turning customer requests into previews in minutes
Speed is often treated as a property of a tool, one that exists separate from its core functionality, but for Goyal, the speed difference “changes the way I interact with Codex compared to other models.”
With Codex, the Braintrust team can integrate iteration into its development workflow rather than letting requests sit and wait. “Codex unlocked our ability to try out customer feature requests in real time,” says Goyal. “Previously, if someone gave us a feature request, it would enter a backlog and get prioritized later.”
Instead, the team can copy and paste requests into Codex, create a preview branch, and show the completed request to the customer in minutes. “The really cool thing with Codex is that we get to iterate and ideate on feature requests with the customer in real time,” says Goyal.
“The more code we write, the more customer problems we can solve, and Codex is the most effective way to do that right now.”
—Ankur Goyal, Founder and CEO
Speed makes autonomous problem-solving possible
For Goyal, Codex changes how much setup it takes to try new ideas. “With other models, I would have to try to prompt the model to solve a specific problem,” he says. Slower tools require more hands-on guidance, which raises the cost of experimentation.
“With Codex, I’ve shifted to writing a test that demonstrates a problem, creating a sandbox environment, and then letting Codex run in that environment,” Goyal says. “This is a novel use case for me, and I can run experiments because of the speed.”
That speed gives the Braintrust team more room to experiment. Instead of prompting step by step, engineers can define the problem, let Codex work in a controlled environment, and move faster from idea to working solution.
“The really cool thing with Codex is that we get to iterate and ideate on feature requests with the customer in real time.”
—Ankur Goyal, Founder and CEO
本文内容采集自官方网站，排版和翻译可能与原页面存在差异。
阅读官方全文