➤ As Claude Code and Codex Get Stronger, What Can AI App Founders Still Build?
This week's Crossing guests are Treg founder Jason Zhou and founding engineer Shi Tianxin. Nine months into their startup, they've pivoted three times — from design agent SuperDesign, to a fully automated growth agent, to Treg, which provides data and tools for agents.
In this episode, we trace their three pivots to explore why vertical agents face a "dead end," why the "zero-person company" is harder than it looks, why AI doing lots of work doesn't necessarily drive growth, and how products like Ego and ChatCut are emerging around coding agents — plus the new bottlenecks humans hit when running multiple agents simultaneously.
New entrepreneurial opportunities are sprouting in the ecosystem of Claude Code and Codex. For a startup team, how do you find the piece worth building, and how do you ride the wave of model improvements? Treg's story offers a live case study.
🎬 Our video podcast will be available on Koji's WeChat Channels, Douyin, Xiaohongshu, Bilibili, YouTube, and other platforms.
📒 The transcript will be published on the CrossingCrossing WeChat official account.
🟢 Part 1 | Three Pivots [00:02:18] Nine months, three pivots: SuperDesign, growth agent, and Treg
[00:03:10] From weekend project to 6,000+ GitHub stars
[00:04:59] Models swallowing applications: Why give up on SuperDesign?
[00:06:59] The "zero-person company" wave rests on a false premise
🟢 Part 2 | Finally, a new direction! [00:14:42] A Codex plugin that unexpectedly changed the growth curve
[00:17:36] Extracting the most valuable layer from the growth agent
[00:18:04] What tools do agents buy most often?
[00:20:13] Data that browsers can't scrape is where the tool market's value lies
[00:21:57] Why is Treg more like Costco?
[00:23:39] The data flywheel of application companies
🟢 Part 3 | Codex / Claude Code are becoming the new platform [00:25:38] Lessons from three pivots: Standing on the shoulders of Claude Code and Codex
[00:26:43] Ego(lite), ChatCut: New tools growing around coding agents
[00:31:56] From Vibe Island to Orca: Opportunities in the multi-agent era
🟢 Part 4 | Humans and agents — how should they actually work together? [00:34:25] A painful pivot decision: No product building for two months
[00:35:01] The "zero-person company" doesn't hold up
[00:36:18] How to do Loop Engineering?
[00:38:50] Running 5–10 sessions simultaneously, humans start becoming the bottleneck
[00:40:13] Claude Code or Codex?
[00:41:41] Grok Bot and cloud sandboxes
[00:43:22] Domestic models, video models, and high-speed models — what scenarios suit each?
🟢 Part 5 | Bury your talent in yesterday, keep building [00:45:08] When programmers can no longer outcode AI
[00:46:05] Life's crossing: From single-skill mastery to cross-domain capability
[00:47:15] The moments they most wanted to quit across three pivots
[00:49:05] Fail Fast: Let doomed directions end as quickly as possible
[00:49:22] If you believe AGI is coming, why start a company at all?
[00:50:28] Even with a universal model, it still needs better tools
Subscribe to Crossing: 🚦 We track the industry shifts and new entrepreneurial opportunities brought by the new wave of AI technology.
🚦 Crossing is Steve Jobs's metaphor for Apple — standing at the intersection of technology and liberal arts, where great products are born. AI is transforming every industry. We seek out, interview, and bring together a new generation of AI founders and active builders in the AI era. Together with them, we explore and embrace the changes and new possibilities.
👦🏻 Host Koji: I founded Crossing, started AI Hacker House — a community space for a new generation of AI founders — and serve as Venture Partner at ZhenFund. I believe technology, especially AI, represents the greatest value-creation opportunity of our generation. Koji on Jike, Koji's website
View this episode's transcript on Xiaoyuzhou