A few weeks ago, an executive was proudly boasting online about a complex integration they had just committed to – with a defunct software vendor. They were bragging about their 10 years of experience doing... the exact same thing.
Here's a tip: If you're bragging about doing something in tech that you did 10 years ago, it might be time for a refresher in new technology.
I was amused. Seriously?! Do people still buy that? I bet their CEO would be interested to know that they just spent millions on obsolete software with flat stock that will be replaced in a few years. Their CEO should talk to me.
This isn't an isolated incident either. I was speaking at LEAP in Riyadh and was surprised by how many outdated dinosaurs showed up at the exhibit hall. Do people still buy this stuff? Snowflake and Databricks weren't there, but the companies they're replacing are making a last stand in the Middle East before their inevitable downfall.
Yup, people still do. I’ll explain why.
To understand why, let me share a personal experience, as I've been guilty of this type of buying behavior myself. Picture an insecure, zit-faced me in a numerical methods class in 2004 at UNR in Reno, trying to learn Mathworks' Matlab. It was so confusing that I had to lean over to the super-nerds in my class and ask for help. How were they making the code work to solve a simple linear equation using matrix math? I didn't like feeling stupid, and I was worried I was drowning. I really struggled during my first numerical class, but that's a common experience when we're introduced to a new programming language or technology.
It's like the hero's journey: we have to go through the valley of despair (incompetency) before we can become heroes. Now, I'm taking the gloves off. I went way beyond being a hero; I became a god.
Fact: I was a Matlab god
I'm serious, no matter who you are, I could outperform you. I'm 100% confident that I was better at Matlab than anyone else. I was creating apps, reading undocumented Matlab as if it were gospel. I did distributed computing for DRI on satellite photo processing, GPU computing, video processing, and nanoparticle image processing. I even compiled Matlab/C++ in production on million-dollar controllers. I deployed Matlab into production process controllers at Intel/Micron and on a 600-GPU cluster for an HPC hedge fund. If you think you can beat me, please share! I was beyond a hero; if you could imagine it, I could build it in Matlab right in front of you in real-time, within 2-4 hours, no problem, shipped. Have 50 people watch me code live – I don't care, have 1000s.
Please forgive my arrogance when it comes to Matlab, but I was a force to be reckoned with. My extreme obsession in Matlab helps the next part of the story where I leave #Godmode, but for the meantime in the world of nerds, I was at the front of the #nerdpack! This was me!!
So, which technology do you think I recommended every time? R? Python? Absolutely not! I didn't want to be at the back of the pack again; that was stressful. I didn't want to feel like an idiot. Matlab was my go-to, the one where I was a force to be reckoned with, the one where I was the best.
It wasn't until I joined HireVue (who I believe gave me a Matlab license – LOL! Sorry, Loren!) that I felt the pressure to switch to Python. When I had to hand my work off to skilled engineers who could code circles around me, I felt the change coming. The licensing costs were prohibitive for their business model, and the engineers were using Python. I saw the language gaining traction in data science back in 2013. I was worried that if I pushed for Matlab, a smart engineer like Brad Grimm, Eli Ribble, or Byron Clark might call me out: "Ben, you're full of it." "Loren, can you fire Ben for being a fraud who can't program in Python like the rest of the industry?" "Loren, I thought you hired a REAL data scientist... Real data scientists KNOW Python..."
So what now? I had two choices: I could go from being a Matlab god at the front of the race to... the village idiot. I didn't want to do that. I searched for every reason to avoid making that change. However, I eventually took the plunge, going from hero to zero and starting the hero's journey anew. I felt like an idiot for months, hating everything. My code was buggy, and I absolutely despised Numpy's matrix notation – it infuriated me. I felt stupid and didn't want anyone to watch me program... ever.
Once I finally made the switch and became average in Python with regards to data science, I realized I was in a better place. Engineers could use my code, it was license-free, and I had access to better algorithms than Mathworks could offer.
Now, let's go back to the industry. Remember the deep-learning library Caffe? It was subpar, and when people began asking for its demise, Caffe users fought tooth and nail to keep it alive. Why? Because of their bias. They were competent, they were heroes like I once was, and their competency made them overlook the fundamental flaws in the language (e.g., having to define a model in a 3,000-line text config file... no thanks).
There are so many outdated languages and software stacks that continue to fight for survival, solely because people have a bias towards their expert-level experience.
Tensorflow? Give me a break. That's a zombie language. I've been saying for years that it would die. Yet, many people still waste their time copying code from it and learning it when they should be focusing on PyTorch instead. MxNet lost ground when it couldn't support a community. Fortunately, the Tensorflow phenomenon is finally fading, although not as quickly as I'd like it to. Just give it a few more years... goooooooooo….. (gone).
This final note isn't for the nerds; it's for the executives.
It's all about replacement – that's it.
I know at least five principal (expert) R users who have converted to Python.
On the other hand, I don't know anyone who has done the reverse – a principal level Python expert who switched to R. So why are your teams burning cycles on R? They should make a strong case that side-steps their obvious bias.
You need to consider who is replacing whom. Who is replacing Snowflake right now? Nobody? Well, why did your VP of Data Science just bet on a platform that is being actively replaced?!
One of the most important metrics to track isn't just the number of customers or revenue, but those metrics divided by customer churn. Why bet on something that's going to be phased out? Well, maybe it's because your people are scared, just like I was. Why on earth would they bet on something they're unfamiliar with? So instead, they choose something that's obsolete, where they're an expert, or better yet, a god of the dead.
*disclaimer this is written on 3 glasses of Caymus - Vineyards, it’s my birthday
Belated happy birthday!!! (Only reading your post now, sorry!)
This post also peaked my interest, because MathWorks (makers of Matlab) is one of my partners. Where is the laughing emoji on substack???