A freelance proposal that opens with their problem, in one paste (the PITCH setup)

This is the setup from the reel. I applied to a freelance job on 31 October: two lines from a client who wanted a ChatGPT newsletter connected to Beehiiv. Between 15 and 20 people had already sent proposals. Most proposals open with "Hi, I'm..." and a list of skills. The client has read that 15 times already.

So the proposal has one job in its first line: show you understood their problem better than they described it.

What you need

• The free ChatGPT or Claude app. Either works.
• The job post, copied in full.
• Three to five true lines about you: what you've done, what you build, one thing you can point to.

Leave out the client's name and anything private from the post. The prompt doesn't need them.

The prompt

Paste this, then the job post, then your lines:

You're helping me write a freelance proposal. Below are the job post and a few facts about me.

Rules:
1. Open with one question or fact that shows I understood THEIR problem, something most applicants would miss. No "Hi, I'm".
2. Then give two concrete ways I'd do it, in plain words.
3. Only use experience from MY FACTS. Never invent clients, numbers or skills. If I'm missing something the job needs, say so to me separately.
4. Under 220 words. End with one question they can answer in one line.

JOB POST:
[paste it here]

MY FACTS:
[3 to 5 true lines about you]

On mine it opened with "ChatGPT can't start itself": a custom GPT only acts while someone is chatting with it, so a daily newsletter needs something outside ChatGPT to start each run. The client never wrote that. It's exactly the thing they'd hit in week one. Then two ways to build it, one fully automatic and one where they read each issue first, and a closing question they could answer in one line.

What it gets wrong

Rule 3 is the important one. It's why, on mine, it added a separate note just for me: my facts said nothing about Beehiiv, the OpenAI API or custom GPT actions, and nothing in them was something the client could click to check. It said so instead of quietly claiming I'd done it.

That's the part to read twice. If the note names something you have done, add one true line about it. If it names something you haven't, don't add it. A proposal that lies gets you a job you then can't do.

It can also be wrong about the client's tools. Mine said Beehiiv may only allow publishing through its API on some plans, and told me to check that before sending, because it couldn't look it up.

The catch

A good first line gets the proposal read. It doesn't get you hired. The client still picks on trust, price and proof, and the prompt can only use the proof you give it. So the three lines you paste matter more than the prompt does.

The measured runs

Two runs, timed, on 31 October 2026, on the real job post (client details removed) with three true lines about me. Not on the phone app: I sent the prompt through Claude from the command line on my desktop, and the times include the tool starting up. The second run is the one recorded in the reel.

• Time: 38 seconds, then 31 seconds
• Length: about 190 and 200 words, each opening with the client's problem ("ChatGPT can't start itself", then "ChatGPT can't publish a daily newsletter by itself")
• Made up: nothing in either. Both listed the gaps instead of filling them

Same prompt, two different proposals. That's normal: run it twice and keep the one that sounds more like you.

No reply from that client yet. If one comes, I'll say so here, and if none does, I'll say that too.

Where the longer version lives

The Rabbit Hole is the community where I post the setups, the ones that worked and the ones that didn't. It's free to join; Premium ($5 a month) adds weekly office hours. skool.com/the-rabbit-hole-3206