My Phone's AI Took Over My Texts for a Day and Somehow Proposed to My Landlord

I want to be clear about one thing before we get into this: I did not mean to tell my mother that she "smells like a discount furniture store after rain." That was the algorithm. The algorithm did that. I am innocent.

Here's the setup. For exactly 24 hours, I surrendered full creative control of every text message, every email reply, and every social media caption to my phone's predictive text and AI suggestion features. No editing. No overrides. If the phone said it, it got sent. The only rule was that I had to let it cook completely before hitting send — no cherry-picking the funny ones and skipping the nuclear ones.

Spoiler: there were a lot of nuclear ones.

The First Hour Was Deceptively Normal

At 8:04 AM, my coworker Dani texted me asking if I'd reviewed the quarterly report. My phone, fully in charge, suggested: "Yes! I looked at it last night and honestly it felt like reading a horoscope written by someone who hates joy."

I stared at my screen for a solid ten seconds. That was... accurate? Uncomfortably so? I sent it. Dani responded with three crying laughing emojis and then "SAME" in all caps. We were off to a promising start.

But the algorithm was just warming up.

When the AI Decided My Friends Needed to Hear Some Truth

By 10 AM, my friend Marcus had texted the group chat about his new apartment, asking what we thought of the photos he'd posted. My phone drafted: "It looks like someone tried to recreate a Pinterest board from memory after one glass of wine. Very ambitious. Very brave."

I sent it before I could process what I was doing.

Marcus replied: "I'm going to need you to explain yourself."

The phone explained itself as follows: "You asked for honesty and honesty is a gift I give freely, like a dog who brings you a dead bird. Lovingly. Aggressively. On your doorstep."

Marcus has since framed this exchange. He says it's the most poetic thing anyone has ever said to him. The algorithm accidentally made a friend for life.

The Landlord Incident (I Am So Sorry, Gerald)

This is the part of the article where I need you to understand that context matters enormously in text messaging, and AI has absolutely no concept of context.

My landlord Gerald texted at 11:30 AM to ask if I was renewing my lease. Standard question. Normal Tuesday stuff. My phone, after presumably scanning my entire text history and deciding it understood human relationships, generated the following response:

"Gerald, I've been thinking about this for a while and I think we should make this permanent. I want to stay here with you forever. You complete my living situation."

I sent it. Gerald responded forty minutes later with: "...I'll draw up the paperwork."

Technically, I renewed my lease. The AI negotiated my housing situation. I don't know how to feel about any of this.

The Social Media Phase Was a Different Kind of Chaos

Around 2 PM, I let the AI write a caption for a photo I posted of my lunch — just a sad desk salad, nothing special. The caption it generated read: "Eating green things because my body is a temple and this temple has clearly angered the gods. Pray for me. Actually don't, I need to sit with this."

That post got 847 likes. My highest-performing content in three years of posting. A machine wrote it in four seconds.

I'm not going to think too hard about what that means for my creative output. Moving on.

What the Algorithm Actually Revealed

Here's the thing nobody tells you about predictive text experiments: they work as a mirror. The AI isn't being creative — it's regurgitating your patterns back at you in slightly scrambled, maximally chaotic form. Every weird phrase it generated was built from words I've actually typed before. The "discount furniture store" mom roast? Apparently I use the phrase "smells like" a lot when complaining about things in my notes app.

The algorithm knows me. It just knows me the way a blender knows a smoothie — all the right ingredients, absolutely no sense of occasion.

The Final Verdict

By 8 PM, I had accidentally insulted two friends (who are now my biggest fans), emotionally proposed to my landlord, and achieved my best engagement numbers of the year on a photo of iceberg lettuce. My mom texted back about the furniture store comment with a string of question marks and then, after an hour, just: "Okay but that's kind of fair."

The AI didn't ruin my relationships. If anything, it stress-tested them and found them surprisingly durable. What it actually did was prove that the version of me living inside my phone's prediction engine is funnier, meaner, more poetic, and considerably more feral than the version of me who second-guesses every text before hitting send.

Maybe the real chaos experiment was the self-awareness we made along the way.

Or maybe I just need a better relationship with my notes app. Either way — Gerald says the lease looks great.