What Stage 1 changed, and why

A tool that learns on its own is only useful if you can see what it learned. Otherwise your messages change, your timing changes, and you find out when the numbers move.

Stage 1 keeps a log. Every change the learning loop makes on its own shows up in one place, with the reason next to it.

Where to find it

Open Insights and tap the tab called What Stage 1 changed. Each entry has a tag for the kind of change, a sentence on what changed and why, and when it happened. The newest is at the top, and the list shows the latest 20.

If nothing's there yet, the screen says so. Changes show up once the numbers can carry them.

What the loop changes

Stage 1 runs a learning pass once a day. It learns from first messages sent by every firm on Stage 1, pooled, and from your own. Three kinds of change land in the log.

Constructs. A construct is a recipe for a first message. Most messages go to the champion, about 15 percent go to constructs in testing, and 5 percent go to the original Hypothesis Opening as a fixed baseline. A construct in testing is promoted when it beats the champion with 95 percent confidence, after at least 100 sends. It's retired when it loses with the same confidence, or when 300 sends pass with no clear answer. It's paused if its opt-outs run at double the usual rate. When a slot opens, the next construct in the queue starts testing. Each of those moves gets a line, and the line says why, like the confidence it won or lost with.

Signal decay. Every signal fades with age, and some fade faster than others. Once a signal family has at least 200 first messages behind it, Stage 1 measures how long until its positive reply rate falls to half of its first week. That becomes its half-life, somewhere between 5 and 60 days. The log names the signal, the new half-life and how many openers it was learned from.

Follow-ups. A follow-up step that has gone out 300 times and earned replies less than half a percent of the time gets turned off for new accounts. The log says which step and shows the replies and sends behind it.

Your own patterns land here too. Once a week, Stage 1 looks at your firm's first messages for what your wins share, or for rewrites of drafts that beat the drafts sent as is. When it finds one, it suggests a private construct from your own messages. The suggestion waits for your approval under Constructs. It never tests until you say so.

You can turn it off

Under Insights, Constructs, the workspace owner can turn Learned changes off. Then pooled data stops adjusting how fast signals fade and which follow-ups go out for your workspace. The tab says when it's off.

Testing has its own switch on the same screen. Turn it off and every first message uses the champion. The workspace owner sets both.

Why it matters for a services sale

A services firm's name rides on every message. You can't let a black box rewrite your outreach and hope it's right. You need to see what moved and why, and you need to be able to say no. That's the same reason a human approves every send.

The log gives you that. When your follow-up schedule gets shorter, you can see the sends behind the call. When a new construct takes over, you can see the confidence it won with. Nothing about your outreach changes quietly.

And it's how the product gets better without you tuning it. Each firm's sends make the pooled numbers sharper for everyone, and your own wins become recipes only your firm uses. More on that in Stage 1 gets sharper the longer you run it.

Every change has a line, and every line has a reason.

Every plan starts with a 14-day free trial and no credit card. Plans and details are on the pricing page, or you can start your trial.