Hi. It's me again. This is going to be more of a tangent than anything. Nothing deep, serious here.
I've been job hunting for almost a year. During that time, I've been targeting specifically startups for 4 months. Also during that time, I've created 3 AI projects (check out my project page). And I must say it's not easy getting a job, especially without those AI projects. Most startups nowadays are AI related so I'm not surprised. It's like a gold rush out there, everyone on the bandwagon. Some startups were delusional, because they just reinvented social media for the n-th time. One even remade Tinder, but AI - Ditto. Some were good. My experience interviewing them was 50/50. I mostly never made it past the screening stage, or "send in the resume" stage. Since most of their job postings were on LinkedIn, I applied on it, but most of my applications were "viewed". That meant they glanced the applications and moved on, which bumped me out. I can understand that I may not have enough, but most of these companies were "vibe-coded" craps selling to clueless venture capitalists anyway. And now they are demanding expertise from you, the applicants ???
I have a couple of stories like this. One time, I was interviewing for an AI firm. These guys came up with using AI agents to help with filling compliance papers, pretty cool and realistic stuff. I researched some of the current tech in preparation for the role. I had an interview with their CTO. It was nice, although I failed to demo my product to him. It crashed my computer halfway (still pissed about that hehe. Screw you, Next.js). I think the thing he was focused on was that I could create an entire app from scratch, run under production, and have actual users. That was why I failed the interview. For me, that is laughable because it's a paradox. If I could build something like that, that it works, and many people would use it, even pay for it, what makes him think I want to work for anybody? I even found an article from San Francisco Standard that shared a similar sentiment.
The next one was from an AI startup for a human-agent collaboration workspace. They wanted the similar skillset as the previous CTO, except the pay is laughably low. The average founding engineer salary I found is around 100k to 180k. They wanted to pay 50k. 50k. For the privilege of working under FAANG co-founders on an app that didn't work, and froze my computer a couple of times. Mind you, my computer is a gaming laptop that can run 8B LLM models. It slowed down a couple of times running those dogshit, "vibe-coded" AI SaaS websites. But this one took the cake. The page caused my computer to freeze for a couple of seconds. I thought I have to restart the machine. Then when I tried to use their product, it didn't work. I just said "Hi" to their AI agent, it just stayed there for a good minute. Brilliant product. This made my app look like a polished experience, because it actually worked, and I only made it in a week with my friend.
Let's change the subject and talk about AI. Because after indulging AI coding agents for a while, I learned something:
- Spec driven development is no different than what I've been doing all this time, except there's a spec document in addition to a technical plan, while the PLAN.md is now a TASKS.md. I haven't used the method, yet. I plan to, once I patch out all remaining bugs in JobbedIn.
- That's another thing. Even with Opus model, when working on either my project or my job, I found my job not moving as fast as I had hoped. Maybe because I didn't attach any browser to the agent, or maybe I need to give more tests and acceptance criterias to the agent. I often still review code by hand, but that I think is for the best. I often catch it giving me lousy code if I don't observe it closely.
I don't hate AI coding as much as I have before. But I am not fully onboard the "vibe" train, either. I think it's a tool, same as a calculator. You still need to know how to do math by hand, just so that you understand the mechanic and can use the calculator more effectively. To use an AI agent, you need to know what exactly you want to do beforehand. That's why to this day, I haven't "vibe-coded" a mobile app, either. I don't understand React Native and mobile development enough to do so.