Maybe you still struggle to write in Pulaar, or in your rinky-dink language. But here in the AI-connected, English-speaking world, we don't need to write anymore. In fact, I don't remember the last time I wrote English. I have one master agent that I talk to and tell what I need to get done. It breaks down the tasks and serves them to spawned agents that get the work done.
I spend $3,000 a month on tokens. I use them mostly for:
- Research and search
- Summarization and transcription
- Writing (the big one)
- Art and music
My great transition this year, as a programmer, was that once I accepted that AI can merge code without needing my review, I could let go. What this meant was that when AI writes English for me, I no longer need to review it before publishing. It can write English, and it writes it well.
Okay, let me add this here before I go too far: /s.
I read this blog post yesterday, and I'm always baffled to hear technical people write these words:
Writing code is solved.
As a software engineer who uses AI, I feel like I must be holding it wrong. I've found a correlation between the number of lines of code written and the number of bugs found. Whether a person or a model generates the code, the longer the PR, the higher the odds of finding an issue. So when someone says coding is solved, I automatically assume they were mediocre at best before using these tools.
If we look at it technically, we could say, in the same vein, that writing English is solved. The AI can write correct sentences, write paragraphs, or publish an entire document. If you have doubts about the content, you can have a second AI verify it and fix potential hallucinations. But of course, that's not how it works in the real world. The moment people detect AI-generated text, even if it is after reading the whole text, they reject it.
It feels empty, deceitful. What is an anecdote from an LLM to convey, when the experience never actually happened? Yes, code is not to be consumed like English. It is for the machine. But it is subjective in its purpose because we humans have to read it eventually.
Writing code is not solved, just as writing English cannot be solved. Writing is subjective and has no solution because it isn't a problem. If anything, AI for writing is a testament to this saying:
"Je n'ai fait celle-ci plus longue que parce que je n'ai pas eu le loisir de la faire plus courte." — Blaise Pascal