AI opportunities · China ↔ Global · Co-founder @F2MLabs
Hong Kong
Joined March 2010
- Tweets759
- Following301
- Followers297
- Likes375
Pinned Tweet
Looking for more partners to build payment infrastructure together. A lot of our recent merchant conversations are with AI short drama and AI companion companies, mostly targeting North America. Many of them fall into high risk verticals, and payment acceptance is a real problem. Mainstream acquirers often won’t take them.
At F2M Labs, we already cover the crypto side. We support Binance Pay through our custodial TSP setup, and recently added a self custody option that brings fees down from around 3% to 1%.
But crypto is only part of the payment stack. These merchants still need card processing and local payment methods. So we’re looking for more licensed acquirers, high risk PSPs and local payment providers that can support these verticals.
If you’re open to partnering, the first two things I’d like to know are simple:
1. Which markets do you cover?
2. Which high risk verticals can you support?
We’re also open to talking with merchants looking for crypto or local payment options. Some of our clients also work with us on TikTok Ads, so traffic and payments often come up in the same conversation.
DM me or leave a comment if this is relevant to what you’re building.
Spent all day at AI Engineer Shanghai 2026. China's AI scene is really buzzing. Both rooms were packed, people were standing to watch talks. Got to chat with Manus, Kimi, HeyGen and World Labs. Really cool, learned a ton.
Ok, can the international political fights just vanish from my timeline for a few days? I wanna see some Phantom Blade discussions😆
Bro, what the fuck? This might be the best tutorial boss I’ve ever played.
Onlyyyyyy Phantom Blade Zero is GOTY.
#PhantomBladeZero
Tweaked a skill all day. Takeaway: AI may know way more than humans, but humans instantly nail the balance between efficient communication and useful info. AI just can’t, especially ChatGPT🤖. Damn it
DeepInfer: faster, cheaper, easier AI inference. Big Blue Fat Fish is handing out the means of production like free pizza.
European firms crying they only make bottle caps. Comrades, now tools are right there.
🚀 Introducing DeepInfer, our in-house inference framework, now fully open source!
Powering DeepSeek-V4.1-Flash at scale with high throughput and low latency. We’re sharing our production stack to help the community lower compute costs, reduce energy use, and make AI more affordable.
🔗 github.com/deepseek-ai/DeepI…
1/4
Jed Zhao retweeted
We gave DeepSeek’s updated V4.1-Flash-1017 the same 100 pull requests with known vulnerabilities and an hour per PR to find them in the @BugBunny_ai VulnPR-100 benchmark.
It found 54 for just $7.53, making it the best-performing open-weight model on our leaderboard.
That matches Grok 4.6’s 54 findings at roughly 16× lower review cost. It also finishes only three findings behind Gemini 4 Argon while costing roughly 22× less.
The previous DeepSeek V4.1 Flash found 44 for $12.79. The update catches 10 more vulnerabilities while costing 41% less, bringing the cost down to just $0.14 per vulnerability found.
Better recall, lower cost, and the strongest open-weight result we’ve measured. Seriously impressive work from the @deepseek_ai team.
There it is. The big blue whale just set a new price performance benchmark. Good luck competing with that.
🚀 DeepSeek-V4.1-Flash-1017 is here!
Stronger benchmark performance, faster task completion, and shorter outputs than the previous version—a Pareto improvement in performance and cost. Now live on our website, app, and API.
🛠️ Optimized for @DeepSeekHarness, Tencent CodeBuddy, WorkBuddy, and more. Upgrade to DeepSeek Harness v0.2.1-rc.1 or later for the best experience with the new model.
1/4
This guy really gave Meta’s shaky AI business a serious boost. At first I thought he was just a rich second-gen immigrant founder. Didn’t expect him to leave his own startup and still have this much impact inside a giant like Meta.
Jed Zhao retweeted
正式介绍开源的 DeepSeek Bot ！（非官方）
DeepSeek Bot ( 简称ds bot ) 是一款基于 DeepSeek Harness ( 简称 dsh ) 的开源插件，让你能够拥有一组持久化的 Bot，使用你自己的 API Key。你可以和你的 Bot 聊天、工作，或者做任何你想做的！
你可以为 ds bot 接入任何你想接入的模型，甚至在未来，你可以直接接入你的本地 agent，这意味着你可以使用自己已有的 ai 订阅。
ds bot 继承了 dsh 的“一切皆插件”理念，未来将会有更多可拓展性，迫不及待想看到会发生什么！
仓库地址：github.com/FeiZhuLulu/DeepSe…
安装：在dsh desktop插件页输入: ds-bot
更多详细信息见评论区↓
After four pretty intense days with 6.1 Sol over China’s National Day break, OpenAI’s taste still feels weak. Opus 5.5 is clearly better. Even with a full UI kit, loose constraints still produce ugly pages.
Luckily my project needs strong, cautious reasoning. That’s still ChatGPT’s biggest advantage.
🚨 codex and claude users !!
6.1-Sol ALMOST performed SAME as Opus 5.5.
Tested 6.1-Sol high and Opus 5.5 high to complete 5 features each.
Sol -> 52% & Opus 42% used of 5-hour limit.
This means a lot !!
Only 10% difference !!
Opus is the top less usage model.
And Sol also can in WITH it too NOW.
So, USE Sol. Not Astra even if you are on Pro or Max.
Comp AI hit $1M ARR in 4 months by turning a $20k+ compliance workflow into $199/mo software. I’m building a legal product on the same logic. This playbook can work in tax, customs, and insurance too.
Spent another whole day working on our legal AI agent, but I almost got into a fight with GPT-6 Sol today. Corrected a few legal mistakes, and it went, “Got it, but I won’t treat that as a general rule.”
WTF 😅 I was like, “It’s literally Article 235 of China’s Civil Procedure Law. It IS the rule in China. Stop making up excuses, or I’m subscribing to Claude instead of Codex.”
Only then did it finally admit it was wrong.
GPT is an idiot when it comes to legal work. I asked it to distill 300K characters of law firm experience. An hour later, it claimed it had extracted everything necessary. Bro, it covered 9 of 35 chapters and ONE of 109 legal documents.
It kept refusing to process the rest. I was laughing in disbelief. Ended up writing a prompt forbidding it from arguing, plus a detailed distillation plan to make it work step by step.