WillHurtDontCare
Legendary Contributor
FASTLANE INSIDER
EPIC CONTRIBUTOR
Read Fastlane!
Read Unscripted!
Summit Attendee
Speedway Pass
Question for programmers out there -
What exactly is "learning by doing" or "moving the needle" in terms of truly understanding a programming language?
I am making great progress on Codecademy (python), but the one thing I've had to face is that everything is right in front of me wrapped up in a neat line for me to do. It gets a bit challenging, but I feel like if I had a blank Pycharm terminal open, implementing these things I'm learning would be difficult.
Just like how business is building product and speaking to users, what is the equivalent of that in terms of learning a programming language without wasting time..?
Is the right way to go to just go through these basic Codecademy courses..?
Is it better in your opinion to grasp a solid understanding of the programming language before building an MVP.. Or is there a better way that allows you to learn faster and get more done as a result instead of waiting months on end to understand how to build the MVP?
Would appreciate some insight. Thanks!
(tagging @WillHurtDontCare , I don't really know any other programmers on here)
With Python, if you're trying to make a web app, start by making a pipeline in Jupyter notebook before you integrate it into a web dev framework like Django. The framework hides what your code actually does behind some user interface that makes it accessible to non-technical people, but that requires adding complexity.
My first "big" consulting project ($20K in 2020) involved webscraping tens of thousands of PDFs, processing them with AWS textract*, loading the data into postgres, and running some regular expression queries to find narrow those 50K or so PDFs, with 10s-100s of pages each, into around 50 pages that actually mattered for the project. I just emailed the client the PDFs, along with some relevant metadata.
However, you don't actually need to create the logic on the back end. In fact, I'm a big advocate of doing as little work as humanly possible because a lot of work has already been done for you. Check out rapid API for lots of useful data, github for lots of useful code, and hugging face for lots of useful and easy to implement machine learning models.
The Important Part of this Post - Market First, Then Program
Don't start with the tech, start with the problem. One mistake that many tech people are guilty of (myself included) is starting with some cool tech and searching for a problem. This is backwards thinking. An entrepreneur starts with a problem that can turn into a business problem that makes money. If people don't have any emotions about whatever you're building, then it won't make money.
What you should do instead is start with where the traffic already goes. You can start with a tool like Mangools ($70 per month) or Spyfu ($39 per month) to find web traffic for particular search terms. Your goal is to find phrases that with high search volume that indicate whether enough people care about your idea to warrant investing time and resources into building a business for it. You can also use Similar Web (free) to see how much monthly traffic your competitors get to determine whether the opportunity is big enough for you to pursue.
The next step would be to get a sense for whether or not you can easily find those people. If you can get email lists of tens of thousands of your ideal prospects, or if you know how to target them with Facebook ads, you might have a business idea. After that, you'd figure out if you can acquire those customers cost effectively. If you find out that your cheapest Google keyword for your service costs $20 per click and you need 10 clicks per conversion for something you charge $25 per month for, you probably don't have a solid business unless you have a pile of cash to wait for them to pay you for several months (disregard this if you have some upsell).
You won't know exactly how much customers cost to acquire until you actually set up the website and collect credit cards, so if you run some napkin math and decide that the numbers are feasible enough for you to test, you make a landing page to test the demand. You don't even need to have a functional product at this point - you can say that your product is in development and collect emails, or ideally offer some deal for people who prepay to get early access. There are too many people who will claim to be interested in your idea until you ask them to hand over their wallet.
After you've verified the demand and that the business is financially viable, just build the pipeline in jupyter locally on your machine. Just get it to do whatever your customers want it to do. Don't even think about offering it in a web app - you can just do things manually then email it to them (that's what I did in the $20K project). If whatever you are offering is valuable enough, the customers will not give a F*ck about how you deliver it. You can pretty up the packaging later.
@mikecarlooch you're a video editing guy, right? There is apparently a huge demand to help people come up with short form content right now, because you can cross post short form content to TikTok, YouTube shorts, Instagram Reels, Twitter, LinkedIn, Facebook video ads, Amazon ads, etc, etc. You could start with something basic like using YouTube's API to find the most replayed sections of the video, then just extract that section of that clip. You could also do something like use OpenAI's whisper transcription to transcribe the video, then find extract whatever clips mention popular keywords. You could find those keywords from Mangools or Spyfu, or you could use some third party API from rapid API to get lists of YouTube comments, Amazon comments, or whatever relevant audience comment source you can get your hands on. You could then feed those comments to GPT 3/4 via the API and ask it what topics people care about. Or you could use spacy to find the most common words, then find the clips that mention those words.
TL;DR on the last paragraph - figure out how to auto generate short form video content from long form video content to post on multiple platforms. I think that Python ffmpeg might help with video clip extraction.
Also, don't try to make the perfect MVP. Just make the shittiest thing that gets the point across to your prospects and hook them with good copy. Then when they hand over money, make the thing great.
Edit: View: https://twitter.com/dafrankel/status/1665504854385541120
*Tesseract is free - if you need to do OCR, use that.
Dislike ads? Remove them and support the forum:
Subscribe to Fastlane Insiders.
Last edited: