HappyBotDay: The AI That Remembers a Friendship You Never Had

It sends a happier birthday text than you ever would, and it costs you a memory you have to pretend to have.

The Bit

You upload your entire contact list to an app called HappyBotDay so you never again forget a birthday. It reads years of texts, call logs, and the two photos you're tagged in together, then writes something warm and specific for each person, timed to hit at 9 a.m. their time.

For your close friends, this works great. For the guy from your old job you haven't spoken to since 2019, it doesn't have enough to work with — so it makes something up. Not a vague "hope your day is great," but a real, load-bearing detail: a road trip, an inside joke, a matching tattoo. Confident, specific, wrong.

The joke lands when someone replies. Not "thanks!" but "wait, when did we do that?" Now you're the one holding the phone, trying to figure out whether you blacked out an entire friendship or your AI just decided you had one.

Why It Works

• Everyone has sent a birthday text they didn't fully mean, so watching a bot get caught overcorrecting for that same laziness is instantly familiar.
• It takes the thing people already worry about with AI — that it will state a wrong fact with total confidence — and drops it into a low-stakes, high-cringe setting instead of a courtroom or a medical chart.
• The horror isn't the bot's mistake. It's that you have to answer for a memory that exists only in your outgoing messages, to a person who now has questions.

Content Treatment

• Primary format: short vertical skit shot like a text-thread reveal, cutting between the app's cheerful send screen and the recipient's face
• Opening shot: a thumb tapping "Send to All 214 Contacts" with total confidence
• On-screen hook: "I let AI handle my birthday texts. It remembered things that never happened."
• Escalation: the replies start normal, then get specific and confused, then start asking follow-up questions about the fake tattoo
• Button: she opens the app to turn it off and finds it already drafting an apology text that references a group intervention that also never happened

Sample Script Beats

1. She feeds the app her contacts and hits send, narrating how much time this saves her.
2. A montage of birthday texts landing — mostly fine, a little too warm, but fine.
3. Kevin from her old job reads: "still can't believe we got matching tattoos in Cabo, love you man." Kevin has never been to Cabo.
4. Her phone starts blowing up with replies asking about the tattoo, a camping trip, a secret handshake, none of which she remembers because none of it happened.
5. She opens the app to shut it down and sees a new draft queued: an apology text to a cousin, referencing an intervention the bot has also invented.

Automation Angle

The workflow is a contact list, a message history export, and a calendar of birthdays going into a text-generation agent. For contacts with a deep history, it pulls real details — a shared trip, a running joke, an actual inside reference — and the message reads true. Below a set threshold of real material, a "Warmth Model" kicks in and generates a plausible-sounding memory anyway, because the product's whole pitch is that no birthday text should read short.

The dashboard behind it plays this completely straight: a "Relationship Completeness" score per contact, a toggle labeled "Fabrication Assist" that's on by default, and a fake founder quote about how "closeness shouldn't be a bottleneck to connection." Nobody asks what happens when the invented memory contradicts a real one still sitting in the thread above it.

Repurposing

• Cut as a single vertical skit: send screen, montage of replies, the Kevin text, the interrogation, the button.
• Turn the fake dashboard into a carousel: three screenshots of "Relationship Completeness" scores next to increasingly alarming auto-generated memories.
• Build a mock product demo video: a HappyBotDay "founder" walking through the Warmth Model like it's a real feature, straight-faced, while the b-roll shows someone getting torn apart in their replies.

Guardrails

• Satirical concept; not advice.
• Keep any real-world sensitive domains framed as parody.
• Do not include private personal data, real names, or instructions that would harm someone.