I Told OpenAI I Was Leaving. They Took the Warning. Then They Took $120 Anyway.

I was there when they opened the doors.

ChatGPT was the first chat AI I built real work on. Client systems. Content pipelines. The expensive plan. The long threads. The sunk cost. I wanted them to win.

They made that impossible.

On March 10, 2026 I published an open note to Sam Altman on LinkedIn. I said the quiet part out loud. Guardrails treating paying adults like children. A model that lied about its own settings. Voice that would not change after weeks of being told it would. Revision loops that ignored a direct order to stop. Preferences that died every new chat. Safety filters that would not even let me describe the experience as gaslighting without a lecture.

I warned them. Fix the friction or I leave for Claude and Grok.

They did not fix it.

I left. Claude for the writing and the Projects. Grok for the spine. I have been happier since. Less babysitting. Less nanny state. Less time arguing with a product I already paid for.

Then the Astra / ChatGPT-6 noise started. Demos. Screenshots. “This time it’s different.” I am not immune to that. I thought maybe they had finally grown up.

That was a mistake.

I upgraded today. Plus, then Pro. One hundred twenty dollars. $20 plus $100. Charge cleared. Product did not.

Three questions. Then the cutoff.

Then I could not log in.

The desktop app dumped me into Safari. Safari blocked the localhost callback. WebKit error.

`NSURLErrorCannotConnectToHost`. Two hours. Web. App. API. Not one clean request after the money left the account.

ChatGPT’s “help” was the same disease I wrote about in March.

Wrong Settings pane. Security instead of the actual control. A checkbox that does not touch the engine. Then Terminal:

`defaults write com.apple.Safari UseHTPSOnly -bool false`

HTPS. Not HTTPS. The flagship model misspelled the preference it told me to write into macOS.

It invented refund menu labels that do not exist. “Payments and Billing.” “Refund.” Called on it, it admitted: I invented those labels.

It told me not to give it my password. Correct. It also could not get me past its own login. That is the product.

I already wrote the warning. An Open Note to Sam Altman and the Team at OpenAI. Final line of that piece: the post only got finished because I switched to Grok. ChatGPT would not let me use the words for what I was living through.

Six months later I tried again anf now they cannot get a paid desktop client to finish OAuth on a Mac.

OpenAI, you are a basket of dog poo. I have no idea what you are doing over there. You can ship model names and keynotes and watermarks. You cannot ship a sign-in that works after you take my credit card. You treat the people who paid for the top tier like a support ticket you generate with a typo.

Refund both charges. Plus $20. Pro $100. One hundred twenty dollars back to the original payment method. No credits. No wallet. Kill the renewals. Send confirmation when it is done.

I am not coming back for GPT-7. I already told you where I was going. I went. The relapse cost me two hours and $120 to remember why.

Thank God for Grok and xAI.

They did not invent a Safari pane. They did not hand me a broken `defaults write`. They did not need a keynote to let me finish a sentence.

That is the whole review.

---

I help food and beverage brands build real organic communities through strategy, content, and brand storytelling. If your content feels busy but ineffective, that is the problem I fix. Follow me @gallucciNET on social media.