Braintrust использует Codex, чтобы превратить запросы на функции в отрасли предварительного просмотра за несколько минут

Braintrust’s engineering team cut the time from feature request to preview branch from days to minutes by running Codex with GPT‑5.5.

Rapid Prototyping with Codex

Codex now turns a customer’s feature description into a working preview branch in minutes, rather than weeks. In one month, 50 % of the Braintrust team migrated to Codex, signaling a company‑wide shift toward model‑driven development. CEO Ankur Goyal notes that Codex “prints more text in the terminal without getting slow, and other models just can’t replicate that.” The speed difference changes how engineers interact with the model, allowing them to copy a request, spin up a sandbox, and generate code in real time.

Speed‑Driven Experimentation

With Codex, Braintrust can iterate on feature requests on the fly. Instead of pushing a request into a backlog, engineers paste the description into Codex, create a preview branch, and show the completed code to the customer within minutes. Goyal emphasizes that this “unlocked our ability to try out customer feature requests in real time.” The model’s rapid output reduces the cost of experimentation: engineers write a test that demonstrates a problem, let Codex run in a controlled environment, and move quickly from idea to working solution.

Scaling the Model Across the Team

The migration to Codex required minimal setup. Engineers no longer need to craft step‑by‑step prompts; they define the problem, let Codex work, and iterate. This shift frees cognitive bandwidth and expands the scope of engineering experiments. By integrating Codex into the development workflow, Braintrust can now prototype new features faster than any traditional coding pipeline.

Real‑Time Customer Collaboration

The most significant benefit, according to Goyal, is the faster feedback loop. Codex enables the team to ideate and iterate with customers in real time, turning abstract requests into tangible code that can be reviewed and refined on the spot. This capability could position Braintrust as the first platform to let customers co‑design AI features during the development cycle.

Looking Ahead

If this pace continues, Braintrust could become the standard for enterprise AI product development, where customer feedback is integrated into the codebase within minutes rather than days. The result: a new era of rapid, customer‑centric AI engineering.

Source: How Braintrust turns customer requests into code with Codex
Domain: openai.com