tux404

I usually go for very brief commit messages, one-liners preferentially. I would just write something on the body when I feel it call for more detailed explanation or if the reason isn't too obvious. I started using plain markdown for the "why" for each project, status, changelog, bugs etc. Every change to the notes is a commit and git became an audit trail with all the dated, uneditable record of all the decisions and changes. Markdown files won it for me because they are the first thing I read when I get back to a project and also because of the ease of access to AI tools.

The caveat is that docs can go stale while commit messages can't, it forces you to be extra careful so to be sure that the files are being correctly updated, I took care of that with a simple script that checks all the notes.

Your concern is legit. Pro is a personal plan under OpenAI's consumer privacy policy: "When you use our services for individuals such as ChatGPT, Sora, or Operator, we may use your content to train our models." As for the API it goes under the business one: "By default, we do not use your business data for training our models."

Claude has the same personal-professional split. Free, pro and Max have trained by default since Anthropic's policy change in August 2025. API and team plans didn't. Choosing between one and another is also choosing a training default. The subscriptions offers the opt-out setting, Data Controls for ChatGPT and Account settings for Claude, and it's worth checking before paying.

I created a similar panel of agent judges each from different companies, to read and analyse AI vendors' privacy policies and terms changes. All must agree and the most cautious label wins, when there's no unanimous decision, I'm the judge.

Recently two bad lines were approved and then published and weren't the judge's fault. As the older changes never went through the panel, an earlier analysis had labeled each change and I then approved those by hand. Two were labeled as no change in meaning and only added one line summaries as notes for the reviewer and not text for the public. I only caught it the next day.

Running the grader once more wouldn't flag it because there was no inconsistency. The fail was human. So my question is when the skill asks whether you'd score a case differently, does it show the text it will publish or only score and reasoning?