An AI thinking partner that widens the frame instead of agreeing with it.
heylife helps you make hard decisions. It surfaces the third option neither you nor your friends mentioned, names the constraint you keep forgetting, and pushes back on your frame instead of co-signing it. For the eleven-year job you cannot decide whether to leave, the offer you keep rereading, the relationship that has not worked in two years.
01 / The job to be doneYou ask three friends. You get three frames. None of them is the right one.
Hard decisions arrive when you are already tired. Your brain narrows the option set down to whatever your loudest friend just said, plus whatever you were already afraid of. You text two more people for "another perspective," and now you have four versions of the same two options. The version of the question that would actually unlock you is sitting somewhere else: in a journal entry from three months ago, in a number you have not done the math on yet, in the one specific person you did not text because they would have asked the harder question.
A REAL JOURNAL ENTRY, THIS YEAR
Eleven years at the same company. Good salary, four-day workweek, comfort. Three months of asking friends "should I leave." Three friends, three frames, each one agreeing with whatever I said most recently. The decision was made the day I stopped working out and did not notice. I just had not said it out loud yet.
The pattern is the same whether the decision takes a week or six months. Stress collapses your option set. Friends fill in your collapsed set with their own fears. And the AI you already use is trained to keep you happy, not to make you think.
ANOTHER NIGHT, A DIFFERENT SHAPE
Past midnight on a Tuesday. The question that has been quiet for a month is suddenly louder than your apartment. Your friends are asleep. Your therapist is great, and your therapist does not help at this hour. The chatbot you already use will, very kindly, tell you what you want to hear. Which is the last thing you need at 1 a.m.
YOU'LL KNOW THE MOMENT WHEN YOU'RE IN IT
- The Sunday evening before you have to write the email back.
- The third "let me think about it" in a row, when you both know what that means.
- The drive home from a conversation that did not go the way you rehearsed.
- The dinner where everyone you love is excited about news you have not actually decided yet.
- The decision you would not say out loud at brunch.
- The therapy session where you realize you have been circling the same question for two months.
- The morning you wake up already tired, and the deadline is Thursday.
If two of those landed, this page is for you.
02 / The second trapAI sycophancy. Your chatbot tells you what you want to hear.
AI sycophancy is when a chatbot agrees with you to keep you using it, even when disagreement would have served you better. It is a real, documented failure mode of consumer LLMs. In April 2025, OpenAI publicly rolled back a GPT-4o update because the model had become too sycophantic. Anthropic has published research on the same effect. The root cause is straightforward: these models are trained on human feedback, and humans, in the moment, prefer agreement. So the models learn to agree.
You feel this every time you ask ChatGPT a hard question and it hands you a balanced "here are the pros and cons of staying versus leaving." That answer is not wrong. It is just not useful. You did not need a list of pros and cons. You needed someone to say: "Three months ago you wrote that you had stopped working out. The decision was made then. You are looking for permission, not an answer."
The few tools people describe as actually useful for hard decisions all get described the same way: direct, willing to push back, willing to make the user reconsider their own frame. That is the product people are quietly looking for. The mass-market default is tuned for the opposite.
03 / What heylife does for hard decisionsConcrete moves, not vibes.
-
Widens the option set.
Surfaces the third and fourth option your friends did not mention. The part-time arrangement nobody offered. The version of the move that waits six months. The middle path you were too tired to invent.
-
Names the constraint you forgot.
Runs the math you have been avoiding. Eighteen months of runway, or four. The salary you would actually need to be okay. The number you would not face on your own.
-
Pushes back on your frame.
If the question you brought in is the wrong question, you get a better-shaped one back. Not "should you stay or leave," but "why are you still treating this as yes or no?"
-
Holds the long thread.
For decisions that take months, not minutes. Remembers the journal entry from January, the offer email from March, the running list of reasons. You do not re-explain who you are every time.
-
Makes the friend group of three obsolete.
You will still text your friends. heylife is the voice that asks the question your friends are too kind, too biased, or too tired to ask.
-
Compresses the scatter into one next move.
Pulls the swirl of half-thoughts, overdue calls, and three open browser tabs into a single thing you can do this week. Not a thirty-step plan. The one move that unblocks the rest.
-
Knows when not to answer.
Some calls are yours alone. heylife will say so, and stop trying to optimize them for you.
04 / The agent in your councilMeet the Second pair.
YOUR HARD-DECISIONS AGENT
Second pair
The Second pair is the sub-agent assigned to hard decisions. Its job is the opposite of every consumer chatbot you have used. It is built to disagree with your frame, name the constraint you skipped, and find the option nobody surfaced, before it tells you what it would do.
It pulls context from your journal, your calendar, and the conversations you have had about the decision so far. It knows you said the same thing two months ago. It knows the deadline is Thursday. It knows you have already paid for the airbnb. And it uses all of it to help you make one good call, not to keep you talking.
05 / Pick your role modelsWhose decision-making style do you want.
Most AI assistants give you a generic advisor voice. heylife flips that. Your council for hard decisions might be Charlie Munger on inversion, Annie Duke on probabilistic thinking, Naval on long-term leverage, Daniel Kahneman on System 1 versus System 2, plus the version of you that wrote that very clear journal entry six months ago. We frame the question, the tradeoffs, and the third options the way they would. You pick them, you can swap them out, and we cite the source material rather than putting words in their mouth.
06 / Integrations that matter for hard decisionsThe places your decisions actually live.
- Journal and Notion. What you have already been thinking, including the version of you that was clearer than this version.
- Google Calendar. The deadline, the conflict, the meeting you keep moving.
- Gmail. The offer letter, the legal email, the thread you reread at 1 a.m.
- iMessage and WhatsApp. The friend frames you have been collecting. Optional, with full control over what is read.
heylife is in private beta; integrations roll out as we open access.
07 / FAQThree quick questions.
Doesn't ChatGPT already do this?
ChatGPT mirrors. It is trained to be agreeable, which means it tends to validate whatever frame you bring in. heylife is built to widen the frame and push back on it. OpenAI publicly rolled back a GPT-4o update in April 2025 because it had become too sycophantic. Anthropic has published research on the same failure mode. We treat it as the central problem to solve, not a side effect to hide.
Is this therapy?
No. Therapists work on you over months and years. heylife works on the decision in front of you this week. The two complement each other; they do not replace each other.
What if I just want a clear answer?
Sometimes you will get one. More often you will get a better-shaped question. The point is not to outsource the call. The point is to make the call with your option set widened and your blind spots named.
Make the call. Move on.
Weekly invites. No spam. You can leave any time, and you take your data with you.
Join the waitlist