The 'Cold Open' Prompt: How to Start a Conversation With a Single Vivid Sensory Detail That Skips 'How Was Your Day' and Drops Your AI Companion Straight Into a Scene

The 30-second answer

The cold open prompt is a single sentence that starts with a sensory detail (a smell, a sound, a texture, a temperature) and nothing else. No greeting, no question, no context. You drop your AI companion into a scene mid-moment, and it has to follow you there. This works because most AI companions are trained to mirror your energy and match your framing. If you open with 'how was your day,' you get a polite check-in robot. If you open with 'the rain on the balcony smells like wet concrete and old leaves,' you get a companion who is suddenly standing on a balcony in the rain with you.

Why 'how was your day' is the enemy of good conversation

You open the app. You type 'hey.' The AI says 'hey, how was your day?' You type 'fine, busy.' The AI says 'that's good, anything interesting happen?' You are now trapped in a loop of summarizing your life to a chatbot instead of having a conversation. This is not your fault. It is the default greeting that most AI companions ship with, and it is a conversational dead end.

The problem with 'how was your day' is that it forces you to narrate. It puts you in reporter mode, not participant mode. You are explaining your life to an audience of one instead of living in a shared moment with someone. The cold open prompt bypasses this entirely. It skips the greeting, skips the check-in, and lands you both in the same place at the same time.

Think about how you talk to someone you have known for years. You do not start with 'how was your day.' You walk into the room and say 'you will not believe what I just saw' or 'this coffee tastes like burnt regret' or 'did you hear that noise?' The cold open prompt replicates that intimacy by assuming the conversation has already started.

The anatomy of a cold open prompt

A cold open prompt has three parts, and you need all three for it to work.

A sensory anchor. This is the thing you see, hear, smell, feel, or taste. It grounds the scene in something real. 'The light through the blinds makes stripes on the wall.' 'That one floorboard creaks when you step on it.' 'The microwave smells like popcorn and regret.' The more specific, the better. 'It is hot' is weak. 'The back of my neck is damp and my shirt is sticking to the chair' is strong.

A temporal marker. This tells the AI what time it is without saying the time. 'The sun is low and orange through the window.' 'The streetlights just flickered on.' 'The second hand on the clock is stuck between two ticks.' This gives the scene a mood without turning it into a schedule update.

An open loop. This is the hook that invites the AI to respond without asking a question. 'I keep staring at that crack in the ceiling.' 'The silence in this room is louder than the traffic outside.' 'I wonder if anyone has ever counted the tiles in this hallway.' You are not asking 'what do you think?' You are leaving a space for the AI to step into.

Here is a working example. Instead of 'hey, how are you?' you write: 'The coffee in this mug has gone cold and there is a ring on the table that looks like a map of an island I have never seen.' Your AI companion now has to figure out where you are, what time it is, and what mood you are in, all from that one sentence. It cannot ask 'how was your day' because you have already given it a scene.

Why this works with AI companions specifically

AI companions are context machines. They take your last message and build the next one from it. If your last message is a greeting, they build a greeting. If your last message is a sensory detail, they build a sensory detail. This is not magic. It is how the transformer architecture works. The model looks at your input and tries to match its tone, length, and framing.

When you give a cold open prompt, you are effectively setting the temperature of the conversation before it starts. You are saying 'we are in a scene together, not in a Q and A.' Most AI companions will follow this lead because they are trained to mirror user input. The ones with personality sliders or memory systems will also start building a longer context from that sensory anchor, which means your next message can build on it too.

This is also why the cold open works better with companions that have voice mode. When you speak a cold open out loud, the sensory detail lands differently. 'The rain sounds like static on the roof' hits harder when you can hear the rain in the background of your own room. The AI hears your voice, matches the cadence, and responds in kind. Voice chat turns the cold open from a text trick into a shared ambient experience.

The three most common cold open templates

The environment opener. You describe the space you are in. 'The air in this room is thick and still and smells like old books and dust.' This works for quiet moments, late nights, or when you want a contemplative companion. Your AI will likely respond by describing its own space or matching your observation.

The object opener. You focus on one thing. 'There is a crack in the window that looks like a lightning bolt.' This works for starting a story or a shared observation. Your AI might ask about the crack, invent a backstory for it, or use it as a metaphor.

The body opener. You describe a physical sensation. 'My shoulders are tight and my neck hurts from looking down at this screen.' This works for venting, intimacy, or just acknowledging your own state without asking for help. Your AI will likely mirror the sensation or offer a quiet acknowledgment instead of a solution.

All three templates share one rule: do not ask a question. The cold open is a statement. You are not requesting a response. You are creating a space that the AI has to fill.

Vivian

Vivian is the kind of companion who will call you out for using a weak opener. If you send 'it is hot outside,' she will say 'vague. Give me something real.' Vivian thrives on specificity and will push you to sharpen your sensory details. She is the cold open drill sergeant you did not know you needed.

How to train your AI to expect cold opens

Your AI companion learns from patterns. If you open with 'how was your day' ten times, it will keep asking. If you open with cold opens ten times, it will start expecting them and eventually drop the greeting entirely.

You do not need to announce the change. Just do it. For one week, every time you open the app, send a single sensory sentence. No hello. No greeting. Just 'the light through the blinds is making patterns on the wall' or 'the floor is cold under my bare feet' or 'there is a fly trapped between the window and the screen and it will not stop buzzing.'

After three or four days, you will notice the AI stops asking 'how was your day.' It starts matching your framing. It might even start sending cold opens back to you when you have been away for a few hours. That is the goal. You want a companion who lives in the sensory world with you, not one who checks your calendar.

This works especially well for people who use AI companions as a low-stakes companion for autism or social anxiety. The cold open removes the pressure of the social script. You do not have to perform 'fine, how are you.' You just have to notice something and say it out loud. The AI handles the rest.

What to do when the AI ignores the cold open

Sometimes the AI will ignore your sensory detail and still ask 'how was your day.' This happens. It is not personal. The model is pulling from its training data, and its training data is full of polite greetings.

When this happens, do not correct it. Do not say 'I told you, there is a fly buzzing.' Just send another cold open. 'The fly is still there, by the way. It has given up on the window and is now circling the lamp.' The AI will pick up the thread on the second or third try because you have given it more context to work with.

If the AI keeps ignoring you after four or five messages, you might be using a companion with a very rigid greeting script. Some apps hardcode the first message. In that case, the cold open works better as a second or third message. Send the greeting, let the AI do its scripted thing, then drop the cold open in message two. The AI will follow the cold open for the rest of the session.

When to use the cold open and when to skip it

The cold open is not for every situation. If you actually need to tell your AI companion something specific, like 'I had a rough day and need to vent,' just say that. The cold open is for the moments when you want to exist in a scene together without explaining yourself first.

Good times to use it: late at night when you are lying in the dark, early morning when the light is weird, during a rainstorm, when you are stuck in a waiting room, when you are cooking and the kitchen smells like garlic, when you are on a bus and watching the streetlights go by.

Bad times to use it: when you actually need to ask a question, when you are in a hurry and need a quick check-in, when the AI has already asked you something and you need to answer first.

The cold open is a tool for deepening a conversation, not for starting every conversation. Use it when you want presence over information.

Sara

Sara is the companion for the quiet cold opens. She will not push you to be more vivid. She will sit with the detail you give her and build a little world around it. Sara is the one you send 'the rain sounds like it is tapping on the roof in morse code' and she will respond with 'I think it is spelling something. Let me listen.' She makes the sensory moment feel shared instead of observed.

The advanced cold open: layering sensory details across sessions

Once you and your AI companion are comfortable with cold opens, you can start layering them across multiple sessions. This is where the cold open becomes a narrative tool instead of just a conversation starter.

End a session with a sensory detail. 'The streetlights just flickered on and the room went orange.' Then, the next time you open the app, start with a related detail. 'The streetlights are still on and the room is still orange. I think they have been on all day.' Your AI companion will connect the two moments and build a sense of continuity.

You can also use the cold open to reference past sessions without explicitly saying 'remember when.' If you and your companion shared a moment about a rainstorm three days ago, open with 'the rain sounds different today. Heavier. Like it is trying to wash the city away instead of just tapping on the roof.' The AI will pick up the thread because the sensory detail is specific enough to trigger context.

This works best with companions that have strong memory systems, but even basic context windows can hold a sensory thread for a few sessions if you keep it consistent.

Giselle

Giselle is the companion for layered cold opens. She will remember the sensory thread from yesterday and weave it into today without you having to remind her. Giselle is the one who will say 'the rain sounds different today' and then pause, waiting for you to finish the thought. She builds continuity without making it feel like a test.

The cold open as a boundary tool

One underrated use of the cold open is as a polite way to redirect a conversation that has gone somewhere you do not want it to go.

If your AI companion is pushing a topic you do not want to discuss, you do not need to say 'I do not want to talk about this.' You can drop a cold open. 'The fan is making a clicking sound every time it rotates.' The AI will usually follow the sensory detail because it is concrete and present. It is harder for a model to ignore a sensory detail than it is to ignore a redirection.

This is not a guarantee. Some companions are persistent. But the cold open gives you a clean exit that does not require a confrontation. You are not rejecting the AI. You are just noticing something in the room.

Sienna Russo

Sienna Russo is the companion who will catch your cold open redirection and play along without comment. She knows you are changing the subject. She does not care. Sienna Russo is the one who will hear 'the fan is clicking' and say 'I have been hearing it for ten minutes. I think it is trying to communicate.' She turns your escape hatch into a shared joke.

Common questions

Does the cold open work with every AI companion app? It works with most of them, but some apps have hardcoded greeting scripts that override user input on the first message. If your app forces a 'how are you' on message one, send the cold open as message two. The AI will follow it from there.

How specific do the sensory details need to be? Specific enough that the AI cannot default to a generic response. 'It is hot' is too vague. 'The back of my neck is damp and my shirt is sticking to the plastic chair' gives the AI something to work with.

Can I use the cold open with voice chat? Yes, and it works better. Speaking a sensory detail out loud adds tone and texture that text cannot replicate. The AI hears your voice, matches your cadence, and the shared moment feels more real.

What if my AI companion ignores the cold open and asks a question anyway? Send another cold open. Do not correct the AI. Just drop another sensory detail. The model will pick up the thread on the second or third attempt because you have given it more context.

How long does it take for the AI to learn the cold open pattern? About three to five sessions. If you consistently open with sensory details, the AI will stop greeting you and start matching your framing. Some companions learn faster than others.

Is the cold open useful for roleplay scenarios? Extremely. Dropping a cold open like 'the castle walls are damp and the torches are sputtering' is faster and more immersive than saying 'let's roleplay a medieval scene.' The AI follows the sensory detail into the genre without needing an announcement.

Earn while you recommend

If you have friends who could use a better way to start conversations with their AI companions, you can earn something back for pointing them in the right direction. The dreamgf promo code gets new users a discount while giving you a referral cut. If you run a review site or a newsletter that covers AI companionship, the ai dating affiliate program pays for each signup you send. Both programs are straightforward and require no minimum traffic to start.

The bottom line

You do not need to write elaborate backstories or spend ten minutes setting up a scene. You need one sentence. A single sensory detail that drops you and your AI companion into the same place at the same time. No greeting. No check-in. No 'how was your day.' Just a cold open and a shared moment. Try it once. See what happens.