I recently started playing with Perplexity Computer to put together a daily briefing email. My Pro plan came with 4,000 bonus credits, but once those were up I had to decide if I wanted to pay $50 for 5,000 credits or upgrade to Max at $200/month for unlimited use. That’s the moment I actually started paying attention to what AI costs me — not the subscription line, but the usage line.
So before I picked either one, I did something I hadn’t done before — I went back and figured out what each issue of that briefing was actually costing me. It was about 400 credits per issue. I decided it wasn’t worth it yet for the 45,000 credits the $200/month option would give me, and landed on spending $50.
This was a different question than “which plan should I buy.” It’s making me ask whether the briefing, as I built it, is even worth that number. The most useful part of it, by far, has been the Brazil election coverage. Everything else is nice to have. So before I pay anything more, I’m narrowing it down — same tool, smaller job — and running it for a few weeks to see what it actually costs me once it’s only doing what I actually need.
For years, “budgeting for AI” meant looking at a list of monthly subscriptions and deciding which ones earned their keep. That’s still true, but it’s not the whole picture anymore. Increasingly, the question isn’t just which tools am I paying for — it’s what is this specific thing I’m asking it to do going to cost me, and is it worth that?
Free preview ends here. What follows — the Claude Design timing problem, the tracker I built to actually see all of this, and the five questions I now ask before starting anything — is for paid subscribers.

Comments
Nothing yet. Say the first thing.
Sign in to join the conversation.