While working in Geremia, the talking skull, I needed to find a simple MP3 player that I could drive from the Arduino Leonardo board. An MP3 player was the simplest solution. A device that is easy to operate via a virtual serial port and not expensive.

After some research, I ordered a couple of DFPlayer Mini for 3.07 US dollars each.

Truth said I destroyed both of them in the first two iterations of the Arduino Proto Shield. No big deal, they are not expensive. Since I wanted to progress in the project, I needed a few more devices as soon as possible. I looked on Amazon and found an offer to buy five for 8.4 US dollars. That price would make 1.68 US dollars for each unit. 

The courier delivered the package; I installed the unit on my proto shield and started playing with the software. You can use many libraries for this MP3 player with Arduino, and the logic is straightforward.

The device is very cheap, and it has limitations. You will have to name the files on the microSD card carefully, ensure there will be no other files apart from MP3 files, and (!!!!) load the files on the card in the same order of their naming.

After half an hour of tinkering, I could play MP3 files on the microSD card.

From a software standpoint, I needed to ensure that the finite-state machine on Arduino was keeping the state BUSY until the specific MP3 file had finished playing. I wrote another proto-thread to monitor the state of the BUSY pin on the DFPlayer Mini. The pin will be low if the player is busy doing something and high if the player is idle.

No way, it was not working at all. Erratic behavior every single time. I checked the wiring, the soldering, and the power levels, but I had yet to find a way to get through. I also connected the device to a logic analyzer to look at the behavior of that pin. I could not find a solution. I searched the net to see if someone else had the same problems and found a solution.

Someone wrote that there are DFPlayer Mini clones with many issues, some of which were similar to those I was experiencing on my device.

Now, wait! A clone for a five-dollar device?

Yes, that’s the case.

After downloading and installing a DFPlayer mini analyzer sketch for Arduino, I had confirmation that the device I was using was a clone. Many of the functionalities that an original device should have were not working on my device.

I had bought cheap clones of a cheap device.

Lesson learned: there will always be a clone hunting you.

Over the weekend, I made some significant progress with Geremia, my talking skull project. I was able to finish the print and assembly of the servo motor mounting bracket, and to my surprise, it worked perfectly. I attached the bracket to the back of the skull and used a short string to connect the servo arm to the jaw. The movement was smooth and responsive, even with the short lever.

One minor issue I need to address is finding a reliable way to make this connection in the final assembly. However, I’m confident that this is a problem I can solve quickly.

After finishing up with the hardware, I returned to the software side of the project. I have been working with two finite-state machines, with one running on the Arduino Leonardo and the other on my MacBook. The PC software is written in Python and communicates with the Arduino board via a serial connection.

The PC software monitors Geremia’s health at specific intervals and checks for new events to be notified via the talking skull. When there is a new notification, the software sends the notification over the serial port to Geremia, which then takes the appropriate action. This includes flashing the two LEDs, moving the skull jaw, and playing an MP3 file specific to the notification type.

The challenge with this setup is that everything needs to run in parallel to make the action appear natural. Unfortunately, Arduino does not support multitasking or multithreading, given the limited hardware resources available. However, there is a solution available – I imported the protothreads library, which simulates multithreading via software.

Modifying my source code to incorporate protothreads was not a significant challenge, although there are a few things to be careful about. For example, you cannot declare local variables in protothreads, so you need to use global variables or static variables inside the proto-thread. Additionally, it’s essential to avoid synchronous functions inside the proto-thread, like using the delay() function. Instead, you need to use PT_SLEEP(), which is part of protothreads.

I have five proto-threads in my application, which include one for the left LED, one for the right LED, one for the servo motor, one for the MP3 player, and one for the serial communication. It’s exciting to see how Geremia’s brain has evolved to be multitasking, even though it wasn’t originally designed to be.

I still need to perform some more testing on the finite-state machines to ensure everything is working as expected. Nonetheless, I’m thrilled with the progress I’ve made so far and can’t wait to see what else I can accomplish with Geremia.

While I was working at the proto-shield development for Geremia, I found myself getting more and more interested in electronics. At the very same time, I found intriguing the intersection between hardware and security.

It was about time to give it a try. 

While trying to sort out the clutter in my drawers, I found an extremely old Netgear WNCE3001 WiFi repeater. It is more than five years old and no longer helpful. What about trying to tear it apart and play with it?

I took it to my desk and started to remove the plastics. That was easy since I had an iFixit kit. I ended up with the motherboard, and a quick inspection revealed that an RTL8196C-GR Realtek microprocessor powered the device and that a nice MX25L3206E CMOS flash was on the other side of the motherboard.

A deeper inspection of the motherboard revealed six pins without any text on the motherboard. Those may have been a JTAG or UART connection.

I picked up my multimeter, powered up the device, and measured the voltage from those pins. I quickly found the ground pin and noticed a varying voltage from another pin. That started to look like a serial interface.

A few months ago, I picked up a second oscilloscope for a little bit more than a hundred bucks and connected one probe to those two pins. After a few seconds of looking at the screen, it was obvious that it was a serial connection.

A few calculations from the scope measurements and I found out it was a 38400 bps serial connection. I connected those two pins to a USB TTL-UART converter and fired up a serial terminal on my Mac. I could quickly look at the boot sequence on my PC screen. A few minutes later, I found the RX pin on the device, and I have a working serial connection between the repeater and my pc.

It was interesting to notice that the boot sequence ended with my terminal in a working shell with root privileges.

I now have a root shell on a Linux box.

I looked around the file system and closely at the boot sequence. In a few minutes, I found my WPA2 key in plain text. 

Even if I knew nothing about hardware hacking and electronics, that was extremely easy to find.

I also wanted to explore the CMOS Flash, trying to dump its content. I had a Buspirate lying around and used it to dump the firmware. The SPI protocol is relatively easy to understand, but, in this case, you don’t need to know it since the Flashrom utility deals with it for you.

I could not read the flash chip while it was on the motherboard. I had to desolder it. No big deal; I would trash the device at the end of the experiment.

After the memory dump, I used binwalk to extract the files from the dump file.

No more exciting things to do with this device.

I could not define this device as the most secure one out there.

Maybe the next time I could try to get one of those cheap security cameras and look at what’s in there.

During the last weekend, I made some steps towards the realization of my talking skull, Geremia.

I completed the client ad server application that will drive the skull. A straightforward state machine that will poll different resources on the Internet and that will notify the Arduino board sitting in the skull to perform actions related to the notification (i.e., blink the LEDs that play the role of eyes, play an MP3 file to simulate speech and activate the servo motor to make the skull jaw move.) The server and the client are connected over a simple RS232 interface.

That was a couple of hours of coding.

It was about time to figure out how to install all of this stuff into the skull. There is enough space to host the Arduino board and the proto shield I assembled. Unfortunately, I could not fit the servo motor, no matter what I tried. There needs to be more space, and giving the servo motor a stable installation would be challenging. I used a DM996 servo motor because the spring holding the jaw is pretty strong, and I needed something with high torque. This is also why I had to implement a voltage regulator circuit on the proto shield while dealing with the Arduino 12 Volt external power supply, but this is another story. 

My first thought was to abandon the moving jaw feature of Geremia. Thinking about it, I realized that it was the coolest feature of the skull, and I didn’t want to drop it simply because I could not find a viable solution. This is when you feel sad because you can’t resort to your father, that was a remarkable mechanical engineer. We could have had some great fun working on this. Again, this is another story.

I started thinking about a possible solution. After some time, I found that the only solution was to host the servo outside the skull, anchored to the skull base.

I grabbed my Remarkable and started drawing, not like a mechanical engineer but more or less like a first grader. After some time, I thought I had a solution. The servo motor would be visible behind the skull, but I can cover it with something. Geremia would look very cool with a fancy scarf.

I have my design ready, but how can I build it? I found that Fusion 360 is free for personal use. I have never used any 3D modeling software in my entire life. I downloaded it and installed it on my Mac. I grabbed a caliper I bought in China decades ago and started to make measurements on the servo motor and the skull base. I added the relevant quotes to my terrible drawing and was ready to model it on Fusion 360. It was mostly made of boxes with four holes to anchor the servo motor.

I viewed a couple of tutorials on YouTube on how to use Fusion 360, and a couple of hours later, I had my design ready.

I saved the file and opened it with Ultimaker Cura to print it on my FLSun Q5 3D printer. I adjusted a few parameters to strengthen the object and launched the print. Sunday morning, the print was finished. I need a few screws and nuts, and I will be able to assemble them with the servo.

While I was having lunch, I was thinking about what I did, and I was surprised. It is incredible what you can do with technology today. Knowledge is available to everyone with an internet connection. Twenty-four hours before, I didn’t know anything about 3D design, and now I have something I designed sitting on my desk. 

This is highly fascinating, at least to me.

I have been very quiet in the last few months—too many things on my desk and too little time to manage them.

Nevertheless, I have made some progress on Geremia, my talking skull.

I have finally assembled a working Arduino Proto Shield to be placed on top of an Arduino Leonardo.

It has not been an easy journey. As I could have imagined, electronics is complex, at least for me. I had to learn many things before putting my hands on the soldering iron.

To name a few, voltage regulators, SPI protocol, power, and ground rails.

I have spent much of my free time learning and am happy with my achievements. Connecting my proto shield to the Arduino Leonardo, uploading my test sketch, and seeing the lights turn on, the servo motor spinning, and MiniDF Player play mp3 files have been an emotion.

From my point of view, software is much easier than hardware. I made at least five tries while building the proto-shield. I understood that trace routing is a fine art in electronics. 

It has been fascinating, though. So fascinating that I ended up with hardware hacking as yet another personal project. I have been messing around with an extremely old wireless repeater. With a few bucks, I bought the necessary hardware and started playing. Extremely fun. Finding the UART and JTAG connection and connecting those to my PC with microprobes was quite fun.

Getting a root prompt from an embedded device is fascinating. I also found out how much insecure these devices are. In a few minutes, and without breaking the device, I got my WiFi WPA2 password.

That said, I have to move to the next step for Geremia. Writing the software. This will be fun, and I feel very comfortable with this. A mix of Arduino code and Python to make Geremia and my PC talk over a serial port.

Hopefully, it will not take a lot of time.

The final step is going to be complicated. Placing the two boards, the eyes, and the speaker, in the skull will be easy. Hosting the servo motor and efficiently connecting it to the jaw will be different. I will design something in Fusion 360 and 3d print it on my FLSun Q5 3D printer.

Anyway, as with everything else, electronics is hard if you don’t know anything about it.

I have always been a great fan of Marco Vichi’s novels. The main character, commissario Bordelli, is intriguing. In one of the last novels, Dr. Diotivede, a pathologist, gives a human skull to Bordelli as a Christmas gift. The skull is then named Geremia and lives in commissario Bordelli’s kitchen. Ok, a little bit weird, I agree.

I could not resist doing the same, so I headed to Amazon and treated myself to a full-sized skull to keep on my desk as a headphone holder. Same name, Geremia.

A couple of weeks ago, I looked at Geremia and said to him, “Hey, you look dead. We should do something about that.”

Since we are in 2023, adding a couple of LEDs for his eyes, an mp3 player to make him speak, and a servo motor to make his jaw move according to the audio file played would be easy. I played a bit with Arduino in the past, which would be the perfect choice for this kind of project. I reached into my drawers and took out everything I already had. I used circuit.io to design the electronic circuit and wrote a state machine on Arduino Sketchin to make Geremia talk with my PC over a serial port. The idea is to use Geremia for notifications.

And then it started. The circuit worked perfectly on the simulator, but I was missing a voltage regulator, a few capacitors and resistors, a good servo powerful enough to move the jaw, and a prototyping board… It started. I have spent too much money on Amazon in the past few weeks to gather all the needed pieces. Now the prototype is sitting on my desk, working.

I will transfer the prototype to an Arduino Proto Board in the next few days. But, wait, I do not have a soldering station… buy one… Well, you get the point; I could not stop.

In the meantime, I browsed YouTube to get some knowledge of electronics. I found a guy who built a 4 bit computer completely on breadboards. I already know which is going to be my next personal project. For the time being, I am having fun reading “Digital Computer Electronics,” which talks just about that.

It is quite funny, but the reality is that it is incredible to realize that social media is not free anymore.

Twitter wants eight bucks a month to let me use MFA and other features that will give me the famous “blue check.”

Facebook and Instagram want twelve bucks a month to give me the “blue check” and more visibility for my posts.

So, at the end of the day, they want all of my personal data and my money as well.

The Internet is becoming a wild place.

I found myself in the position of writing a small command line Python tool for one of my colleagues. Something pretty simple. Get a contacts list from a Google Sheet and create Lead objects on Salesforce. 

Easy stuff. It should not require more than a couple of hours, testing included.

I decide to take a test with this.

I would not write a single line of code.

I describe every single function in the application as a comment and let GitHub CoPilot do the coding for me.

Well, It took twenty minutes to have the application up and running—a few tweaks here and there, mostly on prompts and error messages.

The exciting thing is that those functions are written using my programming style. Honestly, I would have written those functions in the same way. It must be said that GitHub CoPilot has access to all of my private repositories, most of them coded with Python, so it is pretty easy to guess my (not so pythonic) programming style.

I was browsing our Slack channel this morning. I entered one of those and found a reference to useGalileo.ai. 

The basic idea of this service is to write a sentence that describes the UI design you need, and it will provide you with a complete UI design that you can edit later in Figma. 

I guess that out of the box, you could think: “The designer job is going to die!”

Read the previous sentence carefully. There is a verb: think. 

That’s the key.

The most important responsibility of a designer is to think, not to build a nice-looking UI design. That’s the last part of any project, like 10% of the total work. It is pure execution. Be careful; I am not talking about the design language you will use in your UI design. I am talking about asset production.

If there is something that, in a more or less smart way, can cut off production time, that’s welcome. I would be keen to pay for it. 

The reality is that the designer’s work is being able to write that sentence. And that’s where the talent resides. And the money too 😉

I have written a couple of times about why correntedebole.com still exists.

It costs me a few hundred bucks every year. I could find some cheaper hosting, but I am comfortable with who I am right now, and I can also run some personal things over there that are not profoundly exposed. 

From time to time, I have to argue about this approach. These are the most common comments I get:

• It is a waste of money.
• There are dozens of spaces where you can write for free.
• Having a personal blog does not make any sense in 2021, 2022, or 2023…

Yes, it costs me money to run the blog. I need hosting space and an SLL certificate. The point here is that it is my own space. I administer it, and I can run the software I like the most on it. 

Yes, there are dozens of spaces where I can write for free. LinkedIn, Facebook, Medium, and Twitter, just to name a few, where I republish what I write on Corrente Debole. 

There are big disadvantages to doing so:

• Not in every place what I write is going to be mine. 
• If they are free, they probably monetize my writings somehow. I am still fine with the current terms, but these terms may change over time.
• I have to comply with the Terms and Conditions of the site where I write. I have never had issues with those, but again, those terms may change, and I want to be free to publish whatever I want.
• Those free space may change their attitude towards users, and it is their absolute right to do so. Again, I want to be free. Look at what happened on Twitter in recent times.
• Does it make sense to have a blog in 2023? The answer is definitely yes.

So, as far as I will have the money to keep Corrente Debole alive, it will continue to exist.

Being laid off sucks. I have gone through it, and it definitely sucks.

You are being laid off even if you have successfully reached your goals.

I still remember every minute of that day as if it were today.

I remember the people around me not being able to look into my eyes as I walked away. The removal process had already begun in their minds.

I remember driving home with my card boxes, trying to figure out how to tell my significant one I no longer had a job. 

At least it was a human being telling me I was fired. If you consider a General Counsel a human being, he was a human.

I did not sleep that night, but I remember that I got out of bed thinking: “Well, your loss. I know what I am worth. Fuck you.” 

It was tough.

The news spread fast. After a couple of days, I started receiving offers from other companies. I decided to take six months off before thinking about my next move.

I understand how thousands of people feel when they are fired from big tech companies. The numbers are impressive. It is estimated that more than 107.000 people lost their job between 2022 and 2023. 12000 by Alphabet, 18.000 by Amazon, 3740 by Twitter, 9090 by Salesforce, 4100 by Cisco, 11000 by Meta, 10000 by Microsoft, and 3900 by IBM, to name a few. If you are strong enough, you can read the numbers here.

It is interesting to look at how these big tech companies laid off people. Some of them disabled physical access to the workplace, others just fired an e-mail, others held an all-hands meeting, and others locked out people from their work accounts or locked their company’s personal computer.

If you are going to fire thousands of people, is there a “correct” and “fair” way to do it? Every company I cited before has critical assets they need to protect when an employee is fired. Think about it multiplied by thousands.

No, there is no good way to do it.

It is also interesting to notice that while you are firing thousands of people, you are running an event with a concert from a celebrity, or you just announced a 10 billion dollar investment in another company.

Corporate inconsistency.

So, do companies have to have a soul? That is not the case. The distance between how companies represent themselves and how they really are is still abysmal. This is true for every company size.

There must be hundreds of managers that had to make a decision on who and when. Did it hurt? I don’t know. I have been there, and it is the most challenging thing I have ever managed. As I remember the day I was fired, I remember the day I fired someone. 

Being laid off sucks. 

What I can say to those people is to look ahead. I went through it, and it was not easy, but it was the only way.

If I can help any contacts, even with a friendly chat, do not hesitate to get in touch. 

You can detect people’s location in an area with WiFi coverage since years ago. Retailers were looking at it to understand customers’ behavior in their stores. Triangulation over airwaves is not new.

I came across a research paper: Dense Pose From WiFi.

An interesting read.

The basic idea is to determine human pose using inexpensive hardware like WiFi routers.

Here’s what you can do:

Given three WiFi transmitters and three aligned receivers, we can detect and recover dense human pose correspondence in cluttered scenarios with multiple people.

The researcher states that they can do this with a couple of WiFi routers at around 60 US dollars. Interesting.

You can easily detect and monitor people’s behavior using video cameras. It is as cheap as the solution documented by Carnegie Mellon University researchers. The main difference is those video cameras come with more significant privacy problems. People can be identified and tracked over time. Something, I guess, is intended.

The WiFi solution can be considered much more privacy aware than video cameras.

While I was reading the paper, I was impressed by the result. Just look at the pictures in the document to understand the potentiality.

I am impressed by what you can achieve using inexpensive hardware and open-source software. If you have the knowledge, you can build advanced systems that, a few years ago, were available only to big companies and government agencies. 

This is really exciting to me. 

Here’s the link to the original document: Dense Pose From WiFi

Streamlit is a Python package that allows you to build beautiful dashboards from Python without the need for a deep knowledge of front-end development. This is my case. I have never been interested that much in front-end development. I was, and I am a backend developer type.
Since I am playing with different data coming from our systems, I found Streamlit incredibly useful for organizing and visualizing the data I collect. You can run your dashboard locally and deploy it in the cloud when you need to share it with others in your organization.
I love PyCharm, which I use for my Python development.
As for every application, you need some debugging to ensure you are doing the right thing.
When you develop locally, you run a local Streamlit, dedicated web server. How can you debug my script in PyCharm?
I spent some time thinking about it and found a solution that may be helpful to other people out there. This will work in Visual Studio Code as well, but I didn’t try it since I don’t love Visual Studio Code so much.
You can add a Run/Debug configuration to PyCharm like the one you see below:

Now you can set your breakpoints in the script and debug it as you usually do in PyCharm.

I am experimenting with NFC Tags and automation on Home Assistant. I have bought a few writeable NFC tags for these experiments. They are cheap, but I wouldn’t say I like wasting them while testing.

Since Flipper Zero can simulate NFC tags, I decided to use it for my tests.

If you want to do the same, here is what you need to do.

Flipper Zero stores NFC Tag data in a plain text file with a .nfc extension. There is a simple command line python script that allows you to write an NFC tag in the Flipper Zero format. Thank you to evilpete for making this available.

#!/usr/bin/env python
"""
    Generates NFC with URL address data and outputs Flipper NFC "save" file
    this is a 5 min hack, No Guarantees

    ./nfc_gen_url.py https://youtu.be/dQw4w9WgXcQ "Rick Roll" > rickroll.nfc

    requires ndeflib

    Written By: Peter Shipley github.com/evilpete

    From pkg https://github.com/evilpete/flipper_toolbox
"""

import sys

try:
    import ndef
except ImportError as e:
    print("Failed to import ndef")
    print("https://github.com/nfcpy/ndeflib")
    sys.exit(0)

# make sure output is utf-8
sys.stdout.reconfigure(encoding='utf-8')

# for debuging
verbose = 0

nfc_header = """Filetype: Flipper NFC device
Version: 2
# generated with flipper_toolbox
Device type: NTAG215
# UID, ATQA and SAK are common for all formats
UID: 04 10 56 01 74 48 03
ATQA: 44 00
SAK: 00
# Mifare Ultralight specific data
Signature: A5 80 A4 CC A0 C3 A1 F6 8B BE 6F EE 83 A6 B9 EE 36 F8 FB C8 14 5A 23 AA 29 DB 78 56 07 B9 6B 92
Mifare version: 00 04 04 02 01 00 11 03
Counter 0: 0
Tearing 0: 00
Counter 1: 0
Tearing 1: 00
Counter 2: 0
Tearing 2: 00
Pages total: 135
"""

data_static = [
    0x04, 0x10, 0x56, 0xCA,     # 01 serial number
    0x01, 0x74, 0x48, 0x03,     # 02 serial number
    0x3E, 0x48, 0x00, 0x00,     # 03 serial number, internal, lock bytes, lock bytes
    0xE1, 0x10, 0x3E, 0x00,     # 04 Capability Container
]

conf_pages = [
    0x00, 0x00, 0x00, 0xBD,      # 130
    0x04, 0x00, 0x00, 0xFF,      # 131
    0x00, 0x05, 0x00, 0x00,      # 132
    0x00, 0x00, 0x00, 0x00,      # 133
    0x00, 0x00, 0x00, 0x00,      # 134
]


def print_nfc_sub(t_data, file=sys.stdout):

    print(nfc_header, end='', file=file)

    p = 0
    for x in range(0, 540, 4):
        print(f"Page {p}: {t_data[x]:02X} {t_data[x + 1]:02X} "
              f"{t_data[x + 2]:02X} {t_data[x + 3]:02X}", file=file)
        p = p + 1


def gen_nfc_sub(tag_data):

    tag_message = []
    data_list = []

    for x in tag_data:
        if x[0] == 'txt':
            tag_message.append(ndef.TextRecord(x[1], "en"))
        elif x[0] == 'url':
            tag_message.append(ndef.UriRecord(x[1]))

        if verbose:
            print(x, file=sys.stderr)
            print(tag_message[-1:], file=sys.stderr)

    buf = b"".join((ndef.message_encoder(tag_message)))

    m_len = len(buf)

    if verbose:
        print("buf", m_len, hex(m_len), buf, file=sys.stderr)

    data_list.extend(data_static)
    data_list.append(3)         # Message Flags
    if m_len < 255:
        data_list.append(m_len)         # Type Length
    else:
        data_list.append(0xFF)
        lenStr = hex(m_len)[2:].rjust(4, "0")
        data_list.append(int(lenStr[0:2], 16))
        data_list.append(int(lenStr[2:4], 16))
    data_list.extend(list(buf))

    data_list.append(0xFE)      # end of Data

    data_len = len(data_list)

    if verbose:
        print("data_list", data_len, data_list, file=sys.stderr)

    x = 520 - data_len
    data_list.extend([0] * x)
    data_list.extend(conf_pages)

    return data_list


# url_str = "https://youtu.be/dQw4w9WgXcQ"
# title_str = "Rick Roll"

if __name__ == '__main__':

    arg_data = []

    for a in sys.argv[1:]:
        if a.startswith("http"):
            arg_data.append(("url", a))
        else:
            arg_data.append(("txt", a))

    if not arg_data:
        print("requires at least one URL argument\n"
              "Please provide URL and optional Title\n\n"
              f"{sys.argv[0]} : https://some.url  'Title'")
        sys.exit()

    nfc_data = gen_nfc_sub(arg_data)

    print_nfc_sub(nfc_data)

    sys.exit()

Copy the script on your computer (e.g. nfcdata.py) and run it:

python3 nfcdata.py "Test data"

This will create a file that you can upload to your Flipper Zero using the qFlipper desktop application.

You will now be able to emulate the tag on the Flipper Zero.

I am working on a personal project that uses python as the primary programming language. I am definitely in love with Python. Coming from an extensive experience with C, I am very comfortable using it, allowing me to dig into OOP when doing it.

I can use dozen of libraries that make my life easier. Salesforce, Harvest, and Google are easily accessible from any application. For this specific project, I have to deal with Pandas, Numpy and Plotly. I had never used them before.

I confess that everything is new and, sometimes, confusing. 

The documentation for these libraries is humongous, and I thought I could not read it all before starting to write my code. 

The feeling that everything is much more complex than before is overwhelming. Thirty years ago, there was little around programming languages and libraries. Better said, there was a lot, but much less than nowadays. 

You had to be careful about scarce resources. Today the computer I am using to write this post is hugely more powerful than any server I used to work on. I can easily waste resources without running out of them. My programming life is more straightforward.

On the other side, the complexity is increased. You have dozen of options to solve any problem you are facing. I always try to understand what I am doing, but I confess sometimes I Copy and Paste the snippet I need to solve a specific problem. 

I feel guilty when I do that.

Nevertheless, I wanted to complete the task I was working on and see it fast and dirty. I will make it elegant, concise, and unique later on. Truth said, it never happens.

If it works, it will stay the same forever.