Sea and OpenAI’s Codex Hackathon Brings 144 Builders Together in Singapore

The final turnout was more selective than the pre-event expectation of more than 150 developers and over 40 projects, reflecting the event’s limited capacity and application-based selection process.

Why Sea and OpenAI organised the hackathon

Sea and OpenAI presented the initiative as part of a deeper collaboration to support AI innovation and developer communities across Asia-Pacific. The hackathon was designed to let builders learn, experiment and develop practical AI solutions with emerging coding-agent technology.

Its emphasis was on building usable products rather than simply demonstrating a model. Teams were expected to work on applications that could address concrete needs in industries or communities, with Codex integrated into the development process.

Who could apply?

The Singapore event targeted people already comfortable with AI coding tools and production-grade software development. The organisers described it as an event for experienced builders, not an introductory AI workshop.

Teams had to meet these requirements:

• Each team needed three or four members.
• Only 40 team slots were available.
• Applications could be shortlisted based on the quality of the submission.
• Selection considered applicants’ technical backgrounds, proposed build plans and fit with the event’s focus areas.

Applicants were allowed to arrive with an idea, but the application was only the starting point. Teams had to develop their project from scratch during the onsite hackathon.

The three project focus areas

Projects could address one or more of three broad directions:

1. Autonomous and adaptive AI

This category covered agents and applications capable of acting with greater autonomy and adapting to changing or unexpected conditions.

2. AI-native products and operations

This focus area covered products and workflows designed around AI from the outset, rather than adding AI as a superficial feature to an existing process.

3. Deep-domain AI

Deep-domain projects applied AI to specialised industries or narrowly defined areas where domain knowledge is especially important.

Together, the three themes gave teams room to explore both general-purpose AI agents and more specialised tools for particular operational or industry problems.

How Codex and OpenAI support were used

Participants used Codex, OpenAI’s coding agent, to prototype more quickly, improve their development workflows and turn concepts into working applications. The event also included onsite guidance and support from global OpenAI experts, as well as project showcases.

That combination made the event more than a conventional timed coding challenge: teams were expected to show not only what they built, but also how effectively Codex contributed to the development process.

How projects were judged

Judges assessed projects against four main considerations:

• Problem framing: whether the team clearly identified a meaningful problem.
• Build quality: how well the application worked and was implemented.
• Depth of thinking: the strength and sophistication of the team’s approach.
• Use of Codex: how effectively the team used the coding agent while developing the project.

This judging framework placed practical usefulness alongside technical execution. A compelling idea alone was not enough; teams also needed to demonstrate a credible build and thoughtful use of the available tools.

Prizes and participation arrangements

Winning teams were eligible for up to US$30,000 in OpenAI API credits. Members of the top five teams also received ChatGPT Pro subscriptions.

Because the event ran for the full day, every team member was expected to attend throughout the programme. Failure to do so could result in disqualification. Organisers provided meals, tables, chairs and power sockets, while participants were responsible for bringing their own devices and chargers.

Singapore application timeline

The published application schedule was:

• Applications opened: May 11, 2026
• Applications closed: May 28, 2026
• Successful applicants notified: June 1, 2026
• Hackathon: June 6, 2026

With only 40 team places available, the selection process made the quality of the proposed project and the team’s technical preparation important parts of the application.

Where the regional series is headed

Singapore was the opening stop for the series. The initial announcement identified Indonesia, Taiwan and Vietnam as markets expected to follow. After the Singapore event, Sea described the expansion as including Indonesia, Vietnam, Taiwan and Malaysia.

The changing list reflects the series developing after its launch, but the broader direction remained consistent: bring hands-on Codex experimentation and AI-building opportunities to developer communities across Asia-Pacific. Taiwan was subsequently identified as another stop in the continuing regional series.

Why the event matters for AI builders

The Singapore hackathon offered a snapshot of how AI coding agents are being tested in a competitive, product-oriented setting. Teams were not judged only on whether they could generate code; they were asked to define a real problem, build a usable solution in a limited timeframe and explain the role Codex played in getting there.

That emphasis on problem selection, implementation quality and workflow could make the series relevant beyond the winning projects. It positions Codex as part of an end-to-end engineering process—from idea and prototyping through to a demonstrable application—while giving Sea and OpenAI a way to engage directly with builders across the region.