
CTO @BitMEX | ADHD-fueled agentic crack team dictator
New York, NY
Joined December 2011
- Tweets368
- Following247
- Followers618
- Likes439
we will need to rebuild and rewrite much of software to adapt to a new, faster pace.
code used to arrive one drop at a time. soon, if not already, code will be written as if it were a continuous stream, like a violent river twisting and turning as it pleases.
this is not a new thing: big tech cos have had to deal with this with the enormous scale of their engineering teams. but now ai has brought an incredibly elastic and capable workforce wielded by every engineer. this is your problem now, and the only way to adapt is to evolve your systems so agents - even and especially the ones with little context about the right way of doing things - can thrive. make the easy thing the right thing.
your codebase is becoming organic architecture. agent contributions grow around the constraints and guardrails you’ve designed. good software engineering matters even more now! and no, the continuous stream of code is not something you have to watch helplessly. be more ambitious and design systems, frameworks, and languages to control the chaos.
the time for mourning coding is long over
Replying to @josephfcox
Here is Proton CTO's AI manifesto, that I obtained multiple copies of 404media.co/manifestos-fear-…
Quentin Machu retweeted
‼️ BREAKING: US officials believe Nvidia CEO Jensen Huang is the target of a Chinese spy plot.
They believe Alice Huang, a Megaspeed executive seen at his side in Taiwan, is a Chinese intelligence asset.
President Trump has been briefed about the matter.
Quentin Machu retweeted
中国社区逆向拆解了Claude的封号机制。以下是他们的发现
Claude使用积分制。注册时账号获得初始分数,之后根据行为加分或扣分。低于阈值直接封号。回分很难
影响分数的因素:
注册
> 干净IP是最重要的因素。不一定要家宽,但必须干净
> Gmail已经不是最优选,有权重更高的邮箱服务
> 注册时直接弹KYC的号可以扔了
> 用手机App注册会加分
支付
> 刚注册就订阅会扣分
> 直接上Max会大幅扣分。最好等两天先订Pro
> 支付时弹KYC说明账号在考察期,但不一定会封
> 国外实体银行卡加分。Google Play也行。虚拟卡扣分
使用
> 不要一次把限额用完。宁可两个号轮换也不要一个号撸到底
> Desktop客户端有少量加分。尽量少用网页端
> 手机端也有少量加分
> 开启记忆功能加分。记忆文件越多分越高
IP和设备
> 稳定性是关键。不要跳IP,尽量一直用同一个
> 代理必须全局模式并开启TUN
> 多设备登录大幅扣分。理想状态是一台电脑一个客户端
> 频繁换IP、频繁登录、高并发都会大幅扣分
账号年龄
> 老号有巨大的容错空间。2023年的号明显更高
> 分数随时间增长,但很慢
其他
> Windows和macOS没区别。Linux会扣分
> 时区和语言基本不影响。不会因为北京时间就封号
> 跟Claude说自己是中国人没问题,记忆不会触发封号
> 大量使用Fable会扣分
> 升级Max大概率进入观察期,除非是高权重老号
观察期和封号
> 接近阈值时账号可能进入观察期。典型特征是掉登录
> 观察期继续扣分就会封号
> 高分号可以扛过多次观察期
关注我,拆解你每天在用的服务背后到底怎么运作的
Quentin Machu retweeted
Spent the day at Shenzhen RoboTech 2026. Biggest takeaway: the US and China are building completely different layers of robotics.
1/ In the US, everyone's on data collection, world models, sim, teleop software. In Shenzhen, it's all hardware: actuators, motors, sensors, components.
2/ Data collection hardware was everywhere: gloves, head rigs, camera arrays. But I didn't meet anyone actually collecting the data. They build the rigs, others use them.
3/ Lots of humanoid companies, but the serious ones are industrial. The robot boxing drew the biggest crowd and was honestly the least impressive thing there.
4/ Ask any booth "can you make this?" and the answer is always "Give me the CAD, we'll get it done." Lead times run 3 weeks to 3 months.
5/ Small sample runs are welcome, and so are factory visits. Most companies are domestic for now, and nearly all want overseas customers.
6/ Go-to-market couldn't be more different. US hardware sells through LinkedIn, X and events. China sells through Amazon China, Taobao and WeChat.
If you're building robots in the US, the hardware layer you need is sitting here, looking for customers.
Quentin Machu retweeted
Replying to @fortysevenfx
I had the same thing, it showed up in 200+ sessions.
Didn't switch shells though. The agents write bash and zsh doesn't split unquoted vars. One line in ~/.zshenv fixed it:
[[ ! -o interactive && ( -n $CLAUDECODE || -n $CODEX_THREAD_ID ) ]] && setopt sh_word_split no_nomatch no_equals
Interactive zsh stays exactly the same.
Quentin Machu retweeted
Most of these buildings in the Shenzhen science park were built in the last couple years. The park itself is massive.
Quentin Machu retweeted
if you have a qwen3.5 4b around, please try steering neurons L14 N8949 +1 and L13 N7956 -1
now you have evil qwen
recipe: dose × median(‖resid‖ leaving layer L, tokens 1:) / ‖W_down[:, n]‖` at the down_proj at every input
Quentin Machu retweeted
Just spent the last 7 days on the front lines of the vulnpocalypse fending off an active cyber attack on one of our products.
Some thoughts:
- Software factories are going to be required equipment for everyone. There’s just no way to keep up with the pace of highly automated adversaries without one.
- We were lucky because we had spent the last few months getting our software factory running well. It is nice to have for product development but proved to be absolutely essential this week.
- At the minimum you need to have a coding server in the cloud pre-provisioned with read only access to all your logs and all the services you depend on for observability. With that in place you are a prompt away from reasonably good situational awareness, ongoing monitoring with loops, and ability for coding agents to validate fixes against your actual, observed attack patterns.
- Shared coding server also unlocks multiplayer incident response where everyone can see every session and jump in as needed. Sharing session links that can be forked is much higher bandwidth than re-explaining things.
- You don’t need to be Ramp/Stripe/Uber scale to have a working software factory in the cloud. We are a team of 3. Ours is built with an @opencode server (thanks @thdxr!), a forked version of compound engineering optimized for automated workflows (thanks @kieranklaassen!) and a multiplayer agent in slack that handles crons (we have our own harness but pick your favorite here).
- To get our software factory up and running well, we had to stop all new product development for ~6 weeks and just focus on getting the validation systems working that enabled us to reliably shipping PR without human code review. This felt very painful in the moment but was totally worth the eventual speedup in velocity.
- KYC for @AnthropicAI & @OpenAI cyber programs is now easy and quick (at least for defensive access), but a day felt like a lifetime this week. Submit your app now before you really need it.
- Getting Mythos access unlocked a deeper level of analysis that helped us shift from reactive mitigation to proactive patching. Wish we had done this a few weeks earlier. Get started now!
It would behove all Western politicians and entrepreneurs to spend at least a week here. Simply to soak in the ambition and to truly internalize the magnitude of the competition. I'd put the best brains in Europe and America up against anyone, but they have to be set free.
"The pace, and the expectations that came with it, felt almost as jarring [in China] compared with America as America's pace does compared with Europe. Everyone was just on it, all the time." world.hey.com/dhh/seven-year…
Quentin Machu retweeted
TroubleMaker must be the vibiest makerspace I’ve been to. It‘s in Shenzhen‘s tech and electronics district, Huaqiangbei and was started in 2016 by Henk Werner after his hardware startup in Holland went bankrupt.
He thought most Shenzhen makerspaces existed to collect government grants, so he built one with no grants, sponsors or commissions on what members manufacture.
It’s basically a flexible membership space (~$600/month) where foreign builders can come to iterate on their hardware and connect with others doing the same. All kinds of tools and machines are already there and literally any electronics part you need is in a 2 km radius.
We believe voice is the most natural way to interact with intelligence, so we created Dessa, the most authentic full-duplex speech model ever built. Talking with it feels like talking with a person…full of emotion, nuance, pauses, and a natural cadence.
block.xyz/dessa
Quentin Machu retweeted
Yesterday we visited Shenzhen InnoX Academy, a hardware incubator founded by Li Zexiang, often called the godfather of robotics in China. His student Frank Wang started DJI out of his HKUST lab, and Li backed it early.
Most hardware incubators give you access to 3D printers and CNC machines. InnoX also has stress testing rigs for corrosion, drop tests and sustained use, so you find out how your product fails before shipping.
But what I found most interesting is that they help with the first step of manufacturing. Going from a prototype to your first production batch is hard for a lot of hardware startups and where it's easy to lose speed. InnoX has manufacturing people on site who build that first larger batch for you AND help you write the standard operating procedures to go to efficient larger scale production.
They also have a large network of manufacturers, and some of them invest in kind by putting up the capital and the people to make your product. For a hardware startup, that can mean your seed round isn’t only cash but someone providing you with 35 of their workers and factory space to produce your first 100k units.
If you're interested in Shenzhen Robot Valley and want to explore one of the world's fastest-growing robotics ecosystems, visit our website: szrobotvalley.com
Quentin Machu retweeted
DeepSeek V4.1 Flash 1017 runs at 1,521+ tok/s on Fireworks Ultrafast.
Fast enough to ship Tetris in 11 seconds, 7x faster than Claude Opus 5.5.
Join the Ultrafast waitlist: bit.ly/3UA0wKq









































