Cruise Dev

Your ML model works perfectly in your laptop. Then crashes in production. Here are the 6 things you MUST consider before deploying any ML system: (Learn these now or debug them later) 🧵

Nobody is cheaper than someone whose only goal in life is to make money

Don't trust the apology of a man who was caught or is facing the consequences of their action. The truth is, will they apologize if things had gone as they planned?

I decided to quit social media, cos I was getting distracted. I started consuming more than I created. To save myself. I'd be using Buffer to channel my posts. Let's see how this goes!!

2020: Nigerian prince emails = spam 2024: AI-generated personalized emails = spam If your patterns change constantly, hardcoded rules fail. This is where ML wins: - Email spam - Fraud detection - Trend prediction ML learns. Rules don't.

Client: "We need ML in our app." Me: "What problem are you solving?" Client: "Everyone's doing AI..." Me: "That's not a problem. That's FOMO." Here's what I ask instead: - Is the task repetitive? - Do you have data? - Will rules work? Save your money

2016: Google switched Translate to neural ML Result: Translation quality improved more in ONE UPDATE than the previous 10 YEARS combined. This is when ML shines: - Massive scale - Complex patterns - Tons of data - Repetitive task Not every app is Google Translate.

Humans: Show us 3 cat pictures, and we recognize cats forever ML algorithms: Show us 10,000 cat pictures, maybe we'll learn This is why ML only works for REPETITIVE tasks at scale. Not every problem needs ML. Some just need smart humans.

I've seen 30+ ML projects fail in production. Always one of these 3 reasons: 1. Treated ML algorithm as the whole system 2. Didn't plan for data drift 3. Ignored infrastructure costs The model wasn't the problem. The system design was.

When i hit 1k followers i'd celebrate with a cow

At what age did you start coding?

At what point were you about to give up on X?

Consumer ML: 95% → 95.5% accuracy = users don't notice Enterprise ML: 95% → 95.5% accuracy = $millions saved This is why: - Google spends years optimizing 0.1% - Consumer apps focus on speed - Enterprise focuses on precision Know which game you're playing.

@r0ktech 5

what does your current setup look like ?

@dev_TEMITAYO The first generation of tech bros.

Before self taught was discovered...

What else do you do outside of coding or tech?

@kelechisp You need some level of delusion to take the steps others are scared to take.

the world belongs to delusional optimists

@amooh001 Let’s gooo

Promote your x account now 💰

@fidexcode This happens when your non tech friends want you to give them update on how you make money. Bro I’m just reaping 3 years of hard work now. So how can I help you 🥹🥹

It took you three years to learn, but you will teach me the same thing in three days. When I'm not Jesus

Video editing is now a part of my schedule as a solo founder. Never knew this day would come.

@charles_dairo True, the urge to want to add a new feature everytime.

AI makes it easy to build; gluttony makes it hard to launch.