BettaFish: student's 10-day vibe-coded AI thesis tops GitHub, wins ¥30M

BettaFish (微舆) is an open-source multi-agent public-opinion analysis system: it crawls 30+ Chinese and global social platforms, has specialist agents debate findings in a 'forum', and generates structured 30-page research reports from a chat prompt; a follow-up engine, MiroFish, builds simulated 'digital worlds' to predict how events unfold.

One person with AI tools can build a shippable multi-agent product in ten days; social-opinion analysis is the wedge, and the same engine can sell prediction next.

Bootstrapped from under ¥10k of personal savings (Xinhua Finance (新华财经)); then Shanda Group committed ¥30M (≈US$4.2M) to incubate MiroFish after Chen Tianqiao (陈天桥) saw the demo in December 2025.

Summer 2025: Guo Hangjiang (郭航江, aka BaiFu), then a 22-year-old senior at Beijing University of Posts and Telecommunications, wanted to finish his thesis early to free time for an internship. He spent the last ten days of vacation vibe-coding BettaFish (微舆) with Claude Code and released it as an open-source project; a PhD-student mentor dismissed it as 'not a qualified thesis'.

BettaFish crawled to ~1k stars, so Guo wrote a farewell essay on Xiaohongshu titled 'What a 1k-star open-source project gives a student'; a week later tech influencers amplified it and stars jumped from 5k to 30k+ within days, putting the project at #1 on GitHub's global trending chart on 2025-11-03. Emails with internship, partnership and investment offers poured in, and Shanda's founder Chen Tianqiao personally recruited him. Guo moved to Shanghai, built MiroFish (a swarm-intelligence prediction engine) in another ten days, sent Chen a demo on 2025-12-16, and within 24 hours Chen committed ¥30M for deep incubation. MiroFish hit #1 on GitHub Trending on 2026-03-07.

Ongoing: Guo's startup, backed by Shanda's ¥30M, is incubating MiroFish; by March 2026 his team had grown to about ten people mixing AI-native indie developers and senior engineers, and BettaFish had ~42k GitHub stars by August 2026.

Vibe coding compressed a multi-agent product that normally takes a team months into ten days, making a solo student's thesis competitive with funded startups.

The product answered a real, repeated need — enterprises pay teams to produce opinion analysis — instead of being a demo with no buyer.

The virality had a repeatable trigger: a milestone essay that tech influencers amplified pushed stars past GitHub's trending threshold, and the second project re-ran the same play.

A single user complaint ('so what do I do with the report?') redirected the roadmap from analysis to prediction — which is what the investor actually funded.

Shanda paid for the person and the pipeline, not the product: Chen said the tech wasn't outstanding, but the ability to define real problems and ship fast was.

Vibe coding collapses shipping cost, so the scarce skills become defining a real problem and directing AI; the durable bet is turning GitHub stars into paying users.

As of 2026-09-02, Guo's startup operates under Shanda's umbrella, incubating MiroFish with the ¥30M commitment; by March 2026 both projects sat in GitHub's trending top 5 together, and BettaFish had ~42k stars with ~7.6k forks by August. The company is pushing MiroFish toward monetizable decision-support use cases such as financial-market prediction and hot-event simulation, while BettaFish keeps shipping daily iterations in response to community feedback. Whether the bet pays off now depends on converting open-source attention into paid product revenue.