Before you interpret your bill: included usage is split into two pools. The first-party pool covers Auto, Composer 2.5 and Cursor Grok 4.5. Third-party models sit in a separate pool. The permanent 2x applies to the first-party pool only, on self-serve Pro, Pro+, Ultra and Teams on the newer pricing.
So "I switched and my limits doubled" is really "I moved work onto a different budget that happens to be bigger". That is still a win, but it changes how you plan: put the mechanical work (reproduce, grep, patch, write the test) on the first-party pool, and spend the third-party pool on the judgement calls where you are paying for a specific model's reasoning.
Same pool, not the same rate. The pool is drawn down at each model's own token rates, so the two are not interchangeable from a budget point of view.
Cursor lists Grok 4.5 at $2 per million input tokens and $6 per million output, with a fast variant at $4 and $18. Composer sits lower. For the same token volume the more expensive model empties the pool sooner, and the fast variant does it dramatically sooner. Check the usage breakdown in the dashboard after a real day of work rather than assuming parity, because the fast/standard split is where people lose a week of budget without noticing.
Worth adding that the two systems fail differently, and the failure mode may matter more than the arithmetic. A token-metered pool degrades: you watch a number fall and you can slow down or switch models. A time-window limit stops you dead until the window rolls.
If you work to deadlines, a plan you can pace yourself against is worth real money compared to one that can lock you out for hours on the wrong afternoon, even if the raw allowance is smaller.
340 emails is a list. It is not customers, it is not validation, it is a list. I had 1,100 and made $0, and the eventual first sale came from someone who was never on it.
Then your cheapest next test is not another ad round, it is finding where those people already gather and reading how they describe the problem when nobody is selling to them.
402 people saw your actual offer. That is a real sample and it is enough to say something.
If 402 people who searched for this exact thing read what you do and 3 gave you an email, the message is not landing. Could be the page, could be the idea. But "not enough data" is not the excuse here - you have plenty of data about the top of the funnel and none about the bottom, which tells you where to spend the next round.
Pellicles also stop forming when the culture is very old and there's a stack of them in the jar. If you've got four or five layers, peel the bottom ones off and keep the newest.
It's normal for the sheet to look terrible while wet. Judging it before it dries is how people talk themselves into using less glue and getting a bad bond.
Every SaaS page you looked at was written by a company with a marketing team and forty customers. Copying their voice at zero customers reads as costume, and people can tell.
Everyone feels it. It goes away the first time someone replies asking how much it costs and when they can have it, because at that point you are not performing interest in a fake thing - somebody actually wants it and now you have a job.
The discomfort comes back if you leave the page up for a year without building anything. That is the version worth avoiding.
Went back through my notes. Not one of the 12 was currently spending money or meaningful time on this problem. They all agreed it was annoying. That is the whole story and it was in my notes the entire time.
Offered exactly this to the four who were most enthusiastic. Two ignored it, one said they would think about it, one paid. So my real conversion is 1 in 12 and the product is four weeks of my evenings doing it by hand, which is worth knowing before I write any code.
Did you round and back the spine? A flat-back block of 180 pages sewn on tapes is always stiffer than it looks. Rounding gives the sections somewhere to go when the book opens, and backing creates the shoulder the boards hinge against. Without it you get exactly the 90 degree wall you're describing.
Also check the grain hangs straight. If the centre front isn't perpendicular to the floor you can chase bust adjustments forever when the real problem is that the front is riding up.
Airlock or a loose lid? Not because an airlock prevents kahm, it doesn't reliably, but a jar you burp daily gets fresh air across the surface every time, and kahm loves that.
Also decide the threshold before weekend 4 and write it down somewhere with a date on it. Deciding what counts as success after you see the result is how everyone talks themselves into the leave.
Different vote: long stitch. 160gsm is thick, so a coptic means a lot of sections, a lot of holes and a long fiddly sewing session that's easy to abandon halfway. Long stitch is one continuous line, you see mistakes as you make them, and it opens flat enough to draw across if you keep signatures thin. Three folded sheets per signature, not six.
Drop the float test entirely and use volume instead. A bit of starter in a straight-sided jar with a rubber band at the start line. You want peak height and the time it took, not a yes or no.
For future charity shop jackets: buy for the shoulders and the chest, alter everything else. Shoulder width is the one thing that's genuinely difficult and expensive to change, and it's what makes a jacket look wrong when it's off.
Warning from doing this badly once: don't spray it and don't flood it. One thin brushed pass, dry fully, then judge. I went heavy the first time and got a shiny plasticky surface that looks wrong and is now the thing a conservator would have to deal with.
One practical note if you ferment: 3kg of chillies makes a serious amount of gas in the first four days. Use a vessel with real headroom and an airlock, or burp daily, or clean your kitchen ceiling.
Churn maths is the part the spreadsheet hides. $9 individuals churn at 6-8% monthly and you are refilling a leaking bucket forever. $200 companies churn at 1-2% annually once they have put your tool in a process. Same revenue, completely different amount of running to stand still.
Ask specifically what the 45 covers. Some quotes include the hem, in which case doing hems yourself saves you very little and the whole calculation changes.
Third toile with three distinct problems identified and fixed isn't doing it wrong, it's doing it. The people who make things that fit are the people who did the toiles.