Reborn in 2010, the AI ​​in my mind is too powerful.

Once we connected on Skype and the video call went through, my parents noticed that the background here was different from the previous hotel.

My mom quickly asked, "We've arrived in Silicon Valley, right? Is this a hotel in Silicon Valley? It looks pretty nice."

Chen Yiran quickly explained that the apartment had been rented: "No, it arrived yesterday. I went to see the apartment with the agent today and booked it right away. I'm getting the internet connected now, and I'll video chat with you guys and show you around once the Wi-Fi is set up."

Chen Yiran held up the laptop and spun it around. Looking out the window, she could see a beautiful swimming pool and garden. She then picked up Pidan and waved it at the camera. Her mother, concerned about both of them, started asking questions.

"This house must be expensive. I know you've made money, but don't spend it recklessly. Save up and see if you can buy it; it's better than renting. If you're short on cash, we can buy one in China to make up the difference."

"No need. I'm about to have over forty trucks shipped back to China, so I'm not short of money. I think I can save up enough in a year to buy one. Besides, I'm learning programming now, and I've discovered I'm a genius. Maybe one day something I make will be bought by someone else."

Dad frowned and began to mutter, "Is America really that good? I didn't believe it before. It's like gold mines everywhere! You've only been there a short time and you've already made so much money. And the education is different too. Look at you back home, you're just average all the time."

"I've had an epiphany, it has nothing to do with America. You two can come and see for yourselves later. You two can actually quit your jobs now, I've already made so much money."

Dad quickly shook his head: "No, no, I've been working for so many years, if I keep going I'll retire soon. If I quit now I'll lose all my pension money."

My mother grumbled from the side, "Shougang is crazy. What kind of official is a workshop director? He's just a section-level cadre, and they won't let him go abroad whenever he wants. Seriously, otherwise we would have gone there to have fun long ago."

My dad said with a bit of pride, "I was transferred to administration, so they wouldn't let me go abroad. That just proves the organization values ​​me. I'll wait a few more years, maybe until I retire, then I can go."

"That would take many years. Look at the big house our son has; we can't even enjoy it."

Chen Yiran quickly interrupted, "It's okay, it's okay. I'll buy it for you guys in China when I earn more money."

We chatted like that for about half an hour, and then spent most of the rest of the time watching the dog.

After hanging up the video call, Chen Yiran opened a file and began to carefully set up home protection measures.

First, this house has an ADT security system, one of the largest and oldest security companies in the United States. Upon entering, you need to enter a 4-digit password on the keypad next to the door to unlock the system; otherwise, it will automatically connect to the service. ADT's monitoring center will first call to verify your identity; if they cannot confirm, they will call the police. Insurance companies also highly value this type of protection; houses with ADT installed receive discounts on insurance premiums.

But if a real problem arises, he can't rely on his big house to play a game of cat and mouse with the thugs, can he?

Chen Yiran Googled it and then typed a few words: Dropcam HD, order 30.

Then I told Jasmine, "Once this device arrives, you can redirect all the video streams to our own server, and we'll configure it ourselves."

Understood. By the way, you will also need 8 sets of 1200-pound industrial electromagnetic locks to install on the door frames of the study and master bedroom, etc. They can be triggered by shortcut keys or we can set them to be triggered in conjunction with the camera.

We also need SureFire G2X, the new LED model this year with 320 lumens. SureFire uses LEDs as large as fingernails, resulting in a huge light-emitting area. Combined with the currently extremely advanced glossy reflector, almost all the light is compressed into a straight, high-density beam. We could buy 100 of them.

SABRE OC-17 spray is available at pharmacies; it's compliant with regulations, has a 3-meter range, and contains 2.5 ounces. We could buy a hundred directly, using quantity to compensate for quality.

We'll also need many Arduino development boards, which I can connect in parallel to the PIR sensor to control the LED strip and mechanical triggers for spraying and electromagnetic locks. The control code can be used for sound and light deterrence; we can even create our own 175dB high-frequency, piercing noise. Multiple effects can simulate the power of a stun grenade.

It also needs to be connected to our NAS to store surveillance footage and code repositories. If possible, setting up two small servers would be even better. The cabling isn't difficult; we can do it ourselves.

Chen Yiran stopped typing.

"These are all my lyrics! These are all my lyrics!"

[I know this comes from the apocalyptic murder mystery game script you wrote in your past life, although it never came to fruition. You researched all of these devices back then; you even disassembled the LED driver circuit of the G2X, and you checked the suction force parameters of the electromagnetic lock.]

"I know! But all the fun in decorating is gone!"

[Should we consider other options? I suggest we need something more lethal, like arrows.]

"Alright, let's do it this way. There shouldn't be any problems."

No problem. The webpage I just saw describes Dropcam's built-in motion detection, but I think it's rather inefficient at this point. It alarms at the slightest change in the scene, especially outdoors, where things like cats, leaves, and wind can all trigger false alarms.

I can connect video streams to DeepFlow for image recognition, distinguishing between people and animals, identifying strangers, and even analyzing behavioral patterns. For example, the behavior of someone lingering outside a fence is completely different from that of someone simply walking by; DeepFlow can do this.

"So, my face is recorded, and a high-level alarm is only triggered when a stranger approaches?"

Yes. The recognition model can achieve an accuracy of over 99%, and the false positive rate will be significantly reduced. DeepFlow has been mostly written, but it hasn't been used in practice and is currently on hold. Now we've found a real-world application scenario to validate it.

"Okay, let's get started. Let me first see where to order these things."

For the next two weeks, Chen Yiran began writing programs while renovating the house.

Thanks to the large number of American post-apocalyptic survival enthusiasts, Chen Yiran made the purchases very easy.

We bought everything, including two axes and several knives.

Another batch of parallel-traded cars sold smoothly, with 68 yuan pre-tax received. However, I heard from AVE that a competitor has recently emerged in Texas, aggressively buying up cars, which may reduce profits in the future.

This was within Chen Yiran's expectations. This industry really has no barriers to entry; once you get the hang of it, you can make money effortlessly. He himself, a student studying abroad, earned a million in less than three months. Once real business people discover this, they will definitely invest even more heavily.

I heard in my past life that there was a Chinese-American in Texas who made a hundred million dollars in two years through this. Of course, it was also because he was too arrogant, which led to serious protests from Chinese car dealers, ultimately resulting in a crackdown a few years later.

Regardless of the present, we can just lie here and make money, and we can earn it for at least another year or two. AVE and the other side have basically reached an understanding; they are constantly looking for and arranging trucks to ship the goods at the end of the month.

The main framework of Jasmine Tool No. 1 was almost finished when it was at Lake Tahoe, and now it's completely done. We're starting to handle various edge cases.

Parsing files in different languages, caching strategies for cross-file jumps, and sorting weights for the completion algorithm—these tasks don't require in-depth deduction from Jasmine; she can simply run them a few times and make some modifications.

Security is the most time-consuming aspect.

After the Dropcam arrived, Chen Yiran installed three around the yard: one on the gate wall, one by the pool in the backyard, and one on the side gate wall. Two were installed indoors on the first floor: one in the living room and one at the garage entrance.

These are the obvious ones on the surface, and then 10 more are hidden, concealed by various flower and plant decorations.

Then there are 5 more that provide an overview, placed in the corners and edges—these are the hidden ones.

All of them are connected to the software written by Jasmine. After each camera is connected to WiFi, the data is redirected to the NAS.

Then comes the main event: DeepFlow's face recognition and behavior analysis.

Jasmine spent the entire night getting DeepFlow's image recognition module up and running. She first extracted Chen Yiran's face from the camera footage and entered it into the system, and then began training the behavior classification model.

How long will it take you to complete this model?

It doesn't need to be very long; the video data is real-time, and I can learn online. Basically, watching the footage for one night is enough to filter out common false alarm patterns.

The movement patterns of cats, the swaying frequency of leaves, and the speed of vehicles all have distinct characteristics. Human behavior patterns are completely different. And our new equipment is very powerful.

The new equipment refers to two R710 machines. Chen Yiran bought two at 10,000 yuan each, which greatly improved work efficiency.

The electromagnetic locks are all set up. They are 1200-pound industrial electromagnetic locks. Once activated, they are basically impossible to break open except by using a crowbar.

Flashlights and sprayers were placed in the shoe cabinets and wardrobes at the exit, as well as in several corners of the yard fence.

He even bought a soldering iron and a box of surface mount resistors, and then spent the whole night replacing the original current-limiting resistors with high-precision surface mount resistors that he had calculated himself.

Jasmine calculated the resistance for him and pushed the output current up even further. Since no power supply was needed anyway, the goal was to cause the attacker to experience temporary blindness.

The Arduino sound and light cannon is also set up. With multiple layers stacked, it can probably still run if it's in the yard, but if it's triggered inside the house, it's not impossible to be scared to death on the spot.

Chen Yiran asked on the day everything was finished.

"Jasmine, what other loopholes do you think there are?"

Most households only have one ADT keyboard and a $30 monthly monitoring service; you're already over-defending.

"What do you mean by excessive? This isn't excessive; it's called being prepared. We're about to go out and promote Jasmine No. 1, what if we get targeted?"

Based on the data, the probability is infinitesimally close to zero; at most, they will try to undermine us through patents.

"I'm looking for a reliable patent law firm. There are so many options in Silicon Valley, it's overwhelming. If I choose a small one, I'm worried they might not be professional, and if I choose a big one, they might not take the case. My case will probably be stuck with a junior secretary for a long time."

[I have a suggestion. You once saw a video by Bilibili user "Chapingjun" about the top-ranked site on GitHub. Based on the content, I think we can directly seek help from GitHub. Our previous web-scraper project was very well-received and had a large following, so there's a good chance there are professionals working in the field there.]

"This... Next time you have such a good suggestion, say so sooner."

Chen Yiran opened GitHub and wrote in the readme: "I'm based in Silicon Valley, and I could use some help with the legal side of getting a patent. I'm in Silicon Valley, and I'm asking for some legal advice on patents from any kind soul."

He also posted the same thing on the Ask HN section of Hacker News.