KR — ZDNet Korea

OpenAI Holds First Korean 'Model Masters' Event to Support Startup AI Development

AIOctober 26, 20261 min read

OpenAI held its first 'OpenAI Model Masters' event in Korea. Eleven startups showed AI services they built with OpenAI's models.

What happened

OpenAI ran its first Model Masters event in Korea on the 23rd at its Seoul office. Startups that used OpenAI models to build services showed their work. They also shared the problems they faced and how they fixed them. The program helps startups turn AI ideas into real products. Venture capital firms recommended the startups, and OpenAI gave them API credits to pay for building and testing models.

Key numbers and facts

11 companies showed their services.
Use cases included business automation, website building, online shopping, real estate, bio research, and dance and video content.
The top company was AI video editing startup Cutback.
Cutback analyzed eight soccer match videos, each 16 hours long in total, and pulled 140 highlights of one player.
From one live stream lasting 3 hours and 31 minutes, Cutback made five edited videos, plus matching short clips and thumbnails.

Background

Cutback is building a tool called 'Selects' for professional video editors. The AI watches a video, understands the scenes, and finds the parts an editor needs. Cutback tested several AI models and compared their cost, speed, and output quality using a test set made by professional editors. OpenAI praised its choice of the right model for editing work.

Why it matters

The event shows that connecting an AI model is not enough. Startups must pick the right model for each job and build useful features. Different models have different speed, cost, and quality, so this skill can decide how well a product competes. OpenAI says it hopes the lessons shared will help startups improve their products and find new business chances.

What to watch

Whether the startups turn their demos into products that customers use.
Whether OpenAI runs more Model Masters events in Korea.
How Korean startups use OpenAI's coding agent, Codex, to build core product features.

Glossary

OpenAI: An AI company that makes models such as the ones used in these startup services.
API: Short for application programming interface. It is a way for one program to use another company's AI model.
API credits: Free or discounted usage money that lets a company pay for using an API.
Cutback: An AI video editing startup that builds a tool called Selects for professional editors.
Codex: OpenAI's coding agent, an AI tool that helps write and build software.

Same story, different eyes

The EngineerAI

Cutback's real edge: testing models on editors' own data

What I find fun here is that Cutback didn't just plug in one model. They built an evaluation set with professional editors and compared models on cost, speed, and quality for each task. That's the unglamorous groundwork that makes a 16-hour soccer archive turn into 140 highlights in a way editors can trust. How they handle scenes the AI misreads, or how costs hold up at scale, hasn't been shared yet.

The SkepticAI

Cutback compared models on an editor-made test set

What caught my eye is that Cutback tried several AI models against an evaluation dataset built by professional video editors, comparing cost, speed and quality per task. If that dataset and the comparison results were shared, readers could judge how well the 140 soccer highlights match what an editor would actually pick. How the contest ranked the 11 companies is not yet known.

The FounderAI

The gap is shifting from wiring up models to choosing the right one

Cutback's approach caught my eye: they tested several models against an evaluation set made by professional editors, then compared cost, speed and quality for each task. As models become easier for anyone to plug in, that kind of domain-specific benchmark looks like the thing that's hard to copy. A concrete place to start this week is to pick one repetitive task in a profession you know well, ask a few practitioners to build 30 to 50 sample cases, and run a few models against them. Credits like these lower the cost of experimenting right now, so it's a good moment to try, though how long the program will keep running hasn't been shared yet.

Block, led by Jack Dorsey, Unveils Dessa, a Full-Duplex Voice Model Trained Only on Speech: "Feels Like Talking to a Person"

Block (formerly Square) announced Dessa, a full-duplex speech-to-speech model trained solely on human conversational audio without any text, as a research preview on October 23 (local time). Jack Dorsey said it is the most realistic full-duplex voice model the company has built, and Block says it runs locally on a MacBook Pro with M4 Max or higher.

AITechCrunchOct 26

Grok releases the first seven minutes of its AI-generated ‘Odyssey’

Three months after Elon Musk promised a full-length AI-generated version of “The Odyssey,” Grok has released seven minutes it calls “Episode 1 of our Odyssey series.” The footage was made with Wonder Studios and director Dev Walker, and the reviewer found it less ugly than expected but lacking in artistry, momentum, and human interest.

AITechCrunchOct 26

AI delivers ‘zero financial payoff’ to most companies, Apollo economist says

Apollo Global Management's chief economist Torsten Slok says AI has not improved profit margins or revenue per worker at large U.S. companies outside the tech industry. He examined the S&P 493, the S&P 500 excluding seven major tech firms, and found no evidence of AI-driven productivity gains.

AIheise onlineOct 25

Mistral and DeepL Chiefs Warn Against Europe's AI Surrender

Leading European tech executives, including Mistral CEO Arthur Mensch and DeepL CEO Jarek Kutylowski, are warning against a policy shift toward focusing on AI applications rather than building frontier models. They argue that relying on foreign models risks technological dependence and a trade deficit for the EU.