You've read that text eleven times.
"I've changed."
Two words, and somehow you're searching them like they hold a hidden paragraph.
They don't.
A text message can only tell you what someone wants you to believe right now, not what they've actually done.
That gap is the whole problem.
A Text Is a Claim, Not a Record
Reread it as many times as you want.
No new information will appear.
A claim like "I've changed" carries zero details about what changed, when, or for how long it held.
It's a headline with no article underneath.
Compare that to something observable, like showing up on time for three months straight.
One is a sentence about the future.
The other is a pattern you could actually count.
The text feels like evidence because it arrives wrapped in relief, or hope, or the specific voice you missed.
That feeling is real.
The information behind it isn't.
Gestures Have the Same Problem, Just With Better Production Value
A single flower, a surprise visit, a handwritten letter naming exactly what went wrong, these feel heavier than a text.
They're still one data point.
One event, performed once, proves that the person can perform it once.
That's it.
If you're freshly dumped and hoping he'll show you instead of telling you, here's the catch.
Asking him to prove it puts you right back in the position of negotiating your way into the same relationship, before you've seen anything.
So don't ask for a gesture.
Wait for a pattern, and don't let one good night stand in for it.
No Contact Isn't Asking You to Judge Blind
If you're four months into no contact, watching secondhand, here's the actual design of it.
It's not asking you to evaluate character with no information.
It's asking you to wait until there's real information to evaluate.
Right now there isn't any.
"He's doing better," from a mutual friend, is still just words, filtered through a person who isn't you.
It tells you nothing about what "better" means, in what situation, under what pressure.
Treat it as noise, not evidence.
You're not missing a test you should be running.
There's no test to run yet, because there's nothing observable happening in front of you.
That's frustrating, but it's not a flaw in the plan.
It's the plan working correctly.
What Actually Counts, Once You Can See Something
When contact resumes, or you're already back and trying to figure out if this time is different, look for specifics tied directly to what broke things.
Not "better communication."
Something like: he brings up a conflict within a day instead of letting it sit for two weeks.
That's a named behavior, and named behavior can be checked.
If someone's leaning on therapy attendance or a sobriety claim as proof, ask what changed in their week, not their vocabulary.
Sessions attended and days sober are inputs.
The output is what you're actually checking for, things like handling stress without disappearing, or catching resentment before it becomes a fight.
Direct evidence means you saw it happen more than once, in more than one kind of situation, without being asked.
Rumors don't qualify.
One good month doesn't qualify either.
Set the Length of the Test Before You're Standing Inside It
This is where people fold, including people who've folded before.
They decide in advance that words aren't enough, then a real apology or a hard cry arrives and the standard quietly disappears.
The fix isn't willpower in the moment.
It's deciding the length of observation before you're emotionally inside the moment that tempts you to skip it.
Aim for three or four months of the specific behavior, across ordinary weeks and stressful ones.
If it holds under boredom and under pressure, that's a pattern.
If it only held during the part where he was trying to win you back, you'll see that too, eventually, if you actually wait for it.
If your own healing has quietly moved your standards past the point where anything would satisfy you, that's not wasted time.
That's information as well.
It just means the answer was never going to be about him changing enough.