Gondola Used Codex to Ship 25% More Code Without Hiring and Doubled Revenue
The small startup Gondola did not have enough engineers, so sales team ideas sat in the queue for months. Now managers build a working prototype in Codex, and programmers take it to release. Merged code changes per month rose 25% across the team and 50% for the lead engineer, with the same headcount, while annual revenue doubled. If a queue of small tasks is choking your team, have product managers come with a prototype instead of a spec.
AI-processed from OpenAI Blog; edited by Hamidun News
Every small team has the same problem: plenty of ideas, not enough engineers. The sales department brings a good hypothesis, but it sits in the backlog for months because programmers are busy with the core product. Gondola, a platform for creators, solved this by giving the Codex tool from OpenAI to managers as well as engineers. According to the company's data, the numbers are these. The number of changes merged into the main code branch per month rose 25% across the whole team and about 50% for the lead engineer. The engineering headcount did not change. Annual revenue more than doubled, and the company says payroll costs did not increase.
This is a customer case study on OpenAI's own website, so the figures should be read as the company's claims, not as an independent audit. There is still a practical lesson in the story.
What the company is and what the problem was
Gondola was founded by Jared Kleinstein, owner of the production agency Fresh Tape Media. He saw that the authors of successful social media videos rarely got recognition. The platform helps creators show their work, prove its impact, and find clients.
The team is small, and the main constraint was engineers' time. Even small experiments competed with core tasks and lost. Now Codex is used for development, and ChatGPT handles daily routine work.
The "code output" metric here is counted as the number of changes merged into the main branch. It is a rough measure: it shows neither the value nor the quality of those changes. However, the company notes that the changes became larger while pull request review time stayed the same. So review did not become a bottleneck. The number of creators that brands hired through the platform rose 2.5 times.
A prototype instead of a technical specification
The main change happened in the process. Previously, a business team member described a need, wrote requirements, and waited for an engineer to have time for a first version. Now they open Codex Cloud, explore the codebase, put together an interface or scenario, and test the hypothesis. Often a working first version already exists by the time an engineer arrives.
Kleinstein himself and partnerships manager Jojo Hesketh merged their own code into the main branch for the first time in their careers. One example is the Programs tool, where companies search for creators, review responses, manage projects, and track budgets. Hesketh prototyped the scenarios in Codex, and then engineers took them to release. According to Kleinstein, the benefit is that you can quickly build a version, throw away the bad one, and keep the working one. The team started with GPT-5.4, did most of the work on GPT-5.5, and handled refinements and bug fixes on GPT-6.1. The founder estimates the savings in days, if not weeks, of development.
A second example is a map of photographers and videographers in the cities that hosted the football World Cup. The company already had the data, and the window of opportunity was short. The prototype in Codex was put together quickly, engineers took it to launch, and the topic did not go stale. Kleinstein links the speed to a small, clearly bounded amount of work and a hard deadline.
Engineers work with Codex through the command line and editors, and use it on GitHub for pull request review. The tool is connected to repositories, development environments, and Slack so that it has product context. People remain responsible for architecture, security, data handling, and production readiness.
ChatGPT for routine work and what to keep in mind
Outside development, the team uses ChatGPT to prepare proposals, schedules, and to search information in work systems. For example, a morning report pulls data from Gmail, Google Calendar, Sheets, and Drive, Asana, Slack, QuickBooks, and the weather forecast. Doing such an overview by hand, according to the founder, would take "hours and hours." Anything that goes external or affects business decisions is checked by a person.
The caveats are obvious. The tools were not the only reason revenue doubled: the company claims a link between Codex and growth but does not prove it. The success with prototypes is confirmed on small tasks with clear boundaries, such as a map or an internal tool. For complex, high-load systems with strict security requirements, such a process needs to be verified separately.
For developers and product managers, the takeaway is simple. If your queue of small experiments is growing, have product managers bring a working prototype rather than a document. Engineers are left with review, hardening security, and shipping to production. Watch review time: as long as it does not grow, the process holds.
- Nov 21, 2026Amazon Releases SGLang-Neuron: gpt-oss-120b Runs on Trainium Chips Without Changing the API
- Nov 21, 2026Factory Cut Metal Pricing from 4–8 Hours to 2–3 Minutes by Removing AI from the Arithmetic
- Nov 21, 2026Jensen Huang Steps Down as NVIDIA CEO After 33 Years, CFO Colette Kress Becomes Interim CEO
- Nov 21, 2026OpenAI Disbands Robotics Team Again: GPT-6 Astra Completed 74.8% of Tasks in Simulation Versus 51%
Key Questions Answered
What the company is and what the problem was?
Gondola was founded by Jared Kleinstein, owner of the production agency Fresh Tape Media. He saw that the authors of successful social media videos rarely got recognition. The platform helps creators show their work, prove its impact, and find clients. The team is small, and the main constraint was engineers' time. Even small experiments competed with core tasks and lost. Now Codex is used for development, and ChatGPT handles daily routine work.
What are the key details on A prototype instead of a technical specification?
The main change happened in the process. Previously, a business team member described a need, wrote requirements, and waited for an engineer to have time for a first version. Now they open Codex Cloud, explore the codebase, put together an interface or scenario, and test the hypothesis. Often a working first version already exists by the time an engineer arrives. Kleinstein himself and partnerships manager Jojo Hesketh merged their own code into the main branch for the first time in their careers.
What are the key details on ChatGPT for routine work and what to keep in mind?
Outside development, the team uses ChatGPT to prepare proposals, schedules, and to search information in work systems. For example, a morning report pulls data from Gmail, Google Calendar, Sheets, and Drive, Asana, Slack, QuickBooks, and the weather forecast. Doing such an overview by hand, according to the founder, would take "hours and hours." Anything that goes external or affects business decisions is checked by a person. The caveats are obvious.
Build & Ship Production AI Agents Without Boilerplate Code
Start your free 72-hour trial at Hamidun Academy (academy.hamidun.com) — interactive lessons on Claude Code & Cursor, live Tech Radar, and production agent templates.
The AI world, distilled — once a week
Seven stories that actually mattered, hand-picked. No noise, no reposts, no press releases.
Done! Check your inbox for a confirmation.