Approve, don’t build

There is a particular kind of software death that happens on a Tuesday. You bought the thing in February because you were going to post more. You opened it eleven times in March. You have not opened it since.

The tool did nothing wrong. It loaded fast, the design was fine, the features were as described. It just always started with a blank box and a cursor, and the blank box always cost you the one thing you did not have.

The blank box is the product

Most creator tools are, underneath, a text field with an AI attached. The pitch is speed: write your post faster. The reality is that writing the post was never the expensive part. Deciding what to post, remembering to post it, cutting it for four platforms that each want something different, and doing that again on Thursday — that is the expensive part.

So the blank box does not solve the problem. It relocates it. You still have to arrive, decide, and start. The tool just makes the middle 40 seconds quicker.

What we built instead

Stacked has a rule that constrains every screen: you should never start from nothing. Every tool is a set of standing rules that turn sources into drafts. The drafts arrive whether or not you show up. Your job, when you show up, is to read something that already exists and press one of two buttons.

That sounds like a small difference. In practice it changes what the product is for:

• The unit of work is a review, not a creation. Reviews take fifteen seconds. Creations take an evening you were never going to give.
• The tool has to be right often enough to be read, not right always. A draft that is 80% there is a huge win. A blank box that is 80% there is a contradiction.
• The failure mode is visible. If a draft is bad you see it, in your inbox, next to nine good ones. If a blank box is bad you just quietly stop opening it.

The three positions of a tool

Every tool in Stacked sits in one of three positions, and the point of the product is to move it rightwards over time.

1. 1Off. Nothing runs. This is where a tool sits when you have not connected the thing it needs.
2. 2Drafting. It runs on its cadence and puts finished work in your inbox. This is the default, and it is where most people live for the first month.
3. 3Autopilot. It runs and executes without asking. You get told what happened rather than asked what to do.

Nobody starts on autopilot, and we do not push you there. The suggestion appears after you have approved something ten times with fewer than two edits, because that is evidence, and evidence is the only reasonable basis for handing over the keys.

Autopilot is a destination, not a default. A tool that publishes on day one, before you have any reason to trust it, is not saving you work. It is creating a new kind of anxiety.

What never goes on autopilot

This is the part we get asked about most, so it is worth being explicit. Regardless of your settings, these always wait for a human:

• Anything flagged by content safety.
• Anything the model rates below 0.7 confidence.
• Anything involving money — a refund, a discount, a price change, an offer.
• Public criticism, and any reply to it.
• A message to someone who is angry.

Those are the moments where being fast is worth less than being right, and the honest thing is to build the product so it cannot be configured into getting them wrong.

The test we hold ourselves to is not "how fast can you make a post". It is: six weeks after signup, on a Tuesday, does anything good happen without you? That question has a different answer than the one a blank box can give.