1. The night I rewrote a client proposal for the third time

That night I was rewriting a client's proposal — third version, still not happy. I was scrolling Hacker News and spotted a post called "Frog and Toad and the Increasingly Capable Machines." That title hooked me. Ten minutes later I closed my proposal and just kept reading.

2. What the article actually says

The author used to work at Anthropic. Core argument: AI's capability jumps over the past few years have flown under most people's radar, but the boundary between "me" (as the worker) and "the machine" (as the tool) is loosening. This isn't a tutorial — it's a long-form thinking piece.

My friend Xiaolin (she runs a mom-and-baby Xiaohongshu account) messaged me after reading it: "That's literally been my last six months." Her case: she needed a title for a post about "baby refuses solid food." AI spat out five versions in 10 seconds. She spent two hours agonizing over version #6 — because only she gets that specific mom's pain point. I've made this same mistake: when ChatGPT dropped last year I immediately renewed Cursor and bought a pile of tools. 80% of those features? Never touched them.

3. Cost of "copying" this idea

This isn't a tool. It's a blog post.

  • Money: $0 (the blog is free)
  • Time: 20–30 minutes with a cup of tea
  • Tech barrier: knowing 26 letters of the alphabet — the writing isn't dense
  • First step: open frogandtoad.ai in your browser, read from the top

I'm not going to tell you "you must read this" — it's not a tool, you won't die if you skip it. But if you've been wrestling with whether to hand more work to AI, it might resonate.

4. What I'd tell myself at each stage

Just starting out, still finding direction: I'd tell myself to chill first. Your clients aren't even stable yet — what AI can or can't do is irrelevant. Serve the one or two clients you have. Bookmark this post, come back in six months.

Already have 1–2 stable clients: I'd tell myself to read it, especially the "me vs. machine" boundary part. It'll give you permission to dump the grunt work on AI and save your time for client conversations and proposal design — two things AI still can't do well.

Scaling up, want more clients: I'd tell myself to read it carefully and look at your services for parts that can split into "AI does + I polish." The author's "don't overestimate, don't underestimate" reminder hits hard at the scaling stage — it'll save you from dropping big money on tools you'll never use.