The Experiment: We curate our Instagram feeds to project a specific image: Success, Happiness, "Living My Best Life." But what does an emotionless algorithm see? I fed my last 12 Instagram posts into GPT-4 Vision and asked it: "Roast this person. Tell me what their insecurities are based on these photos. Be mean."

> ROAST 01: THE "WORK HUSTLE" PHOTO

IMG_2024_01.JPG TYPE: Laptop + Coffee Cup
Photo of a MacBook in a coffee shop with code on the screen and a latte art heart. Caption: "Grinding on a Saturday."
"Ah, the classic 'I need validation for working' post. You have carefully positioned the coffee cup so the logo is visible but the code is blurry because you're probably just changing CSS colors. This photo screams: 'I have no hobbies, so I made my job my personality.' You aren't grinding; you are just lonely in public."

Damage Report: Ouch. Accurate. I was indeed just changing CSS colors.

> ROAST 02: THE "NATURE" PHOTO

IMG_2024_02.JPG TYPE: Hiking Landscape
A wide shot of a mountain view. Me looking away from the camera, candidly. Caption: "Disconnecting."
"Caption says 'Disconnecting,' yet you hiked up a mountain, staged a 'candid' photo (which means you asked a friend to take 10 versions), edited the saturation, and uploaded it to a server farm instantly. You didn't disconnect; you just changed your background wallpaper for the audience."

> ROAST 03: THE "GROUP DINNER" PHOTO

IMG_2024_05.JPG TYPE: Friends holding drinks
Group selfie at a bar. Everyone smiling too hard. Flash on.
"Look at the frozen smiles. This looks less like a fun night out and more like a hostage video where the ransom demand is 'Likes.' You are standing slightly in front to angle your jawline. Insecurity detected: You are terrified of becoming the 'funny friend' instead of the 'hot friend.' Spoiler: You are already the funny friend."

> THE PSYCHOLOGICAL PROFILE

After analyzing the whole feed, I asked for a summary.

> SUBJECT ANALYSIS: "THE PERFORMER"

Core Motivation: Fear of Irrelevance.

Visual Patterns:

  • Avoids direct eye contact in photos (Simulated Depth).
  • Uses humor in captions to deflect from genuine vulnerability.
  • Posts heavily on Sundays (The "Scaries" validation loop).

Conclusion: This user wants to appear "effortlessly cool" but the data shows high effort and moderate coolness.

> THE REACTION

Reading this actually hurt. It wasn't just mean insults; it was data-driven insults. It noticed patterns I didn't verify myself. The "looking away" thing? I didn't realize I did that in every photo until now.

We think we are unique. The AI sees us as a collection of cliches. It recognized the "Latte Art Hustle" trope instantly because it has seen 50 million other people post the exact same photo.

> CONCLUSION

If you want to feel humble, let an AI roast you. Your mom lies to you. Your friends lie to you. The algorithm just processes the pixels.

I deleted the "Grinding" photo. It was right. It was cringe.