YuPeng Wu

Cofounder at FloatMiracle.

[YuPeng Wu @RocNeverFish]

@johnjyang @paperboy_ai Wechat QR code has expired. Plz update!

[YuPeng Wu @RocNeverFish]

Recently I’ve been doing an unhealthy amount of vibe coding.

One thing I didn’t expect:

AI is already very good at generating pages.
But the real pain is all the “small” capabilities around it.
Sidebar.
Selection translate.
Floating toolbar.
Drag resize.
Command menu.
Mobile edge cases.

They look simple, but they’re surprisingly hard to make stable.
Now I’m starting to think:
the future moat may not be code generation itself,
but battle-tested capability layers for AI-native builders.

[YuPeng Wu @RocNeverFish]

I used to describe Git Hired as a small test:

“What kind of AI-native builder are you?”

But I think the bigger idea is different.

LeetCode helped developers build reputation through coding challenges.

But AI-native work is not just coding.

It’s how you:
- frame ambiguous problems
- use AI to move faster
- turn ideas into artifacts
- collaborate with people
- ship visible progress

I’m exploring whether Git Hired can become a reputation layer for AI-native builders.

A place where candidates prove how they work,
and companies discover signals beyond resumes.

Still very early.

But this feels like a more important problem than just another quiz.

[YuPeng Wu @RocNeverFish]

I built a small test around this idea:
https://t.co/LfKzqZ8FCb

It’s rough, but I’ll keep improving it in public.

Would love feedback:
does this feel useful, or still too abstract?

[YuPeng Wu @RocNeverFish]

AI is making job titles feel less useful.

“Designer”, “PM”, “Engineer”, “Marketer” — these labels don’t fully explain how someone actually creates value with AI.

So I’m exploring a different question:

What’s your AI-native work style?

[Teamo @teamo_agent]

🚀 We built Teamo Router to cut OpenClaw API costs by 50%.

Same models (Opus4.6, Sonnet, GPT-5, Gemini). Half the price. 💰
🔑 One API key. ⚡ Installs in 30 seconds.

Join our Discord for early access 👇 https://t.co/siUAUa804q

[YuPeng Wu @RocNeverFish]

Built for OpenClaw power users who are tired of:
❌ API cost chaos
❌ Juggling multiple keys
❌ Paying full price for every model

If that's you — we're doing personal onboarding for the first 100 users in Discord:
https://t.co/QGUFcE1d8F

[YuPeng Wu @RocNeverFish]

If you use OpenClaw with multiple AI models, your API setup is probably a mess.

We built TeamoRouter to fix that.

[YuPeng Wu @RocNeverFish]

TeamoRouter = one API key for all of them.

OpenAI, Anthropic, Google, DeepSeek, MiniMax — all routed through one endpoint.

~50% cheaper than direct API pricing.

Just tell your agent:
Read https://t.co/C5XIvvoFak and follow the instructions

[YuPeng Wu @RocNeverFish]

@tarekgoesplaces @sufeheisenberg @huggingface :) Thanks for this great feedback, would you please merge this info to our official feedback space on https://t.co/Mlt9Yaxfqk

[YuPeng Wu @RocNeverFish]

@tarekgoesplaces @sufeheisenberg @huggingface Nice counter!

[biyang guo @GuoBiyang]

Thanks, AK! @_akhaliq We have put in a lot of hard work over the last 40 days. Now we put forward the first Human-ChatGPT comparison corpus (HC3), open-source the first set of ChatGPT detectors. For more details, please refer to our research paper: https://t.co/PKRUYtPMHc

[Xin Zhang | 张鑫 @xinzhangai]

Our datasets & models are released @huggingface, paper @arxiv:
How Close is ChatGPT to Human Experts? Comparison Corpus, Evaluation, and Detection
https://t.co/JzaLWvLfiL

The paper is still rough, valuable comments are welcome!
#ChatGPT #GPT3 #chatbot

[YuPeng Wu @RocNeverFish]

Plus:
Good news for MODEL FANS and DATA SEEKERS!
The correspondong #model and #data has been opensourced on🤗

Important:
Now you can download world first chinese chatgpt dataset on https://t.co/tyPG8fufkG
and a brand new english chatgpt data on https://t.co/R7HFuLUm14

[YuPeng Wu @RocNeverFish]

Our paper is now avaliable!🤪
My favourite is the discussion about the difference of #ChatGPT and human reviews.

​https://t.co/4rmQVzXUSL

We realize there are still much work to do. 🤣
Check this out!

[Heisenberg @sufeheisenberg]

After a month of effort, we launched the first bilingual (EN/ZH) #ChatGPT detecting tool set, consisting of three different models! 🎉

Detectors on 🤗
@huggingface
: https://t.co/Y9xZEUtwLM

Models/dataset/paper will be released at: https://t.co/BLD5QfNjkb
#ChatGPT #C3Dataset

[biyang guo @GuoBiyang]

After a month of effort, we launched the first bilingual (EN/ZH) #ChatGPT detecting tool set, consisting of three different models! 🎉

Detectors on 🤗
@huggingface
: https://t.co/sTApiyGF4n

Models/dataset/paper will be released at: https://t.co/AKdZ7hEafp
@_akhaliq #GPT3

[Xin Zhang | 张鑫 @xinzhangai]

We launched the first version of our ChatGPT detectors! 🎉

Models, dataset & paper will be released at: https://t.co/MoU3W7IoI3

Try our detectors at @huggingface ❤️!
One of three types: https://t.co/ITfFWIaPJx

Both English & Chinese are supported.

#ChatGPT #C3Dataset

[YuPeng Wu @RocNeverFish]

No Cheating with #ChatGPT!🧨
Our ChatGPT detector support both English and (probably 1st one which supports) #Chinese !😎

Models, dataset & paper will be released in about a week~🥳

Three different detectors are avaliable on @huggingface !
https://t.co/NT659GgPMx