You can’t count how many times people mention Skynet (yep, from Terminator) whenever AI makes a leap forward.

Here’s what a developer thinks about it.

Is AI going to replace us?

Well, AI is already replacing us.

Lots of developers have already plugged AI tools into their daily workflow. This lets them do things like:

  • Automate boring, repetitive tasks
  • Swap out Google for more specialized agents or platforms
  • Optimize and refactor existing code
  • Make debugging easier

The benefits can be huge:

  • Save time
  • Save energy
  • Boost creativity

For a long time, I thought AI would only handle “low-value” tasks.

I pictured a little helper taking care of the dull stuff, but recent breakthroughs show AI can do way more.

What are LLMs?

LLMs, or large language models, are the hot topic right now. Everyone’s racing to build the best one to win big because this market looks like a classic winner takes all scenario.

Some of the popular ones are Claude and Mistral, but there are tons more.

There are online platforms where you can use and especially train these models with instructions (those famous prompts) and user data.

Same deal here, there are literally hundreds, but deepcoder is doing well.

Now you can quickly generate entire apps just by giving the right prompts.

This is a major leap because ChatGPT and other generic platforms used to be limited here, especially on free plans. You’d only get cut-down scripts or half-baked code snippets with little real use.

AI for reverse engineering

AI’s progress in reverse engineering is also mind-blowing-and this is a big deal.

Lots of proprietary apps compile and lock down their source code.

For your average tinkerer, rebuilding that code manually is tough and requires serious reverse engineering skills.

But AI is really good at disassembly, transpiling (turning code from one language into another), and deobfuscation.

Deobfuscation goes beyond just “unminifying” a minified JS script-it makes code readable that was meant to be unreadable.

Of course, you need the right prompts and to supervise the AI, but it speeds things up a lot.

This mostly makes it easier to copy apps and, ultimately, existing features.

AI gives the illusion that everything’s easy

For now, you still need real expertise to write good prompts and fix the AI’s screw-ups.

These platforms can confidently spit out false info (the so-called hallucinations).

If you can’t tell the good stuff from the nonsense, you’ll waste tons of time without getting what you want.

Code itself has never really had value

In my humble opinion, code by itself isn’t that valuable, AI or not. It’s just the engine for features.

Even with strict licenses, there have always been abuses.

But AI speeds up the process and multiplies possibilities, which explains the current feeling of being overwhelmed.