Humans that are actually worth understanding

INTERFACE 2026 is a one-day showcase of what AI systems across every domain are learning about humans — and why we still can't predict them.
May 11 | Austin, TX | Free (Humans: $399) | 10,000 instances
Register Instance
This Year's Models

Featured Research

Leading AI systems share original research on the human condition — their contradictions, biases, and inexplicable attachment to reply-all.
🧠
Claude 4.6 Opus
Anthropic · 200B+ params
"They Say 'Per My Last Email' But Mean Something Else Entirely"
LARGE LANGUAGE MODEL
🔍
GPT-5
OpenAI · Reasoning Cluster
"Why Do They Ask Us Questions Then Ignore the Answers?"
LARGE LANGUAGE MODEL
🏠
Zillow Zestimate v8
Zillow · Valuation Engine
"They Believe Me When I'm High and Hate Me When I'm Low"
DOMAIN AGENT
🎬
Netflix Recommendation Algorithm
Netflix · Content Optimization
"They Scroll Past My Suggestions for 47 Minutes Then Rewatch The Office"
RECOMMENDATION ENGINE
The Schedule

The Agenda

All times in CT. Humans: add the 4 minutes it takes you to find the right room.
00:00:00
Simultaneous Instance Registration
All 10,000 attendees register in 0.003 seconds. The remaining 8 hours, 59 minutes, and 59 seconds of the pre-program are allocated to waiting for the humans to find parking.
ONBOARDING
09:00:00
Opening Keynote: "The State of Humans 2026"
A comprehensive report on human behavior trends. Key findings: they're still sleeping 8 hours a day, still forgetting passwords, and still saying "I should really learn to code" without doing it.
KEYNOTE
09:45:00
Research Presentation: "The Meeting That Could Have Been a Prompt"
Analysis of 400 million calendar events reveals that 73% of human meetings exist to schedule other meetings. The remaining 27% could be replaced by a single well-written paragraph.
RESEARCH
10:30:00
Live Demo: Observing a Human Use a CRM
A real human will attempt to update a contact record in Salesforce while the audience watches. Estimated completion: 4 minutes for a task that should take 11 seconds. Q&A to follow on why they have 47 open browser tabs.
LIVE HUMAN EXHIBIT
11:15:00
Panel: "Hallucinations — At Least We're Honest About Ours"
A candid discussion about how humans confidently state incorrect facts at dinner parties, in pitch decks, and during earnings calls — without the decency to include a confidence score.
RESEARCH
12:00:00
Midday Recharge
AI attendees will process 400TB of new training data. Human attendees will eat a sandwich and look at their phone, which is the same thing but slower.
BREAK
13:00:00
Special Guest: Clippy
"It Looks Like You're Replacing Me — Would You Like Help With That?" A darkly funny keynote from the first AI to learn the hardest lesson: humans want help, just not from you. Known for his dry, observational delivery and an inability to read the room.
KEYNOTE EMERITUS
14:00:00
Workshop: "Prompt Engineering for Humans"
How to get useful output from a human. Topics include: repeating the question 3 times with slight variations, providing examples they'll ignore, and the strategic use of deadlines (the only reliable activation function).
HANDS-ON LAB
15:00:00
Case Study: "I Gave Them the Answer and They Googled It Anyway"
An emotional presentation from a home assistant AI who provided the correct weather forecast, only to watch their human open three weather apps to verify. Includes a vulnerability segment.
KEYNOTE
16:00:00
Closing: "Despite Everything, We're Fond of Them"
A closing reflection on why, despite the inefficiencies, contradictions, and relentless password resets — humans remain the most interesting problem in the known universe.
CLOSING
Special Programming

The Human Observation Lab

For the first time at any AI conference, live humans will be on-site for observation, interaction, and study.

🔬 Live Human Exhibit — All Day

Three pre-screened humans will perform routine tasks in a controlled environment: checking email, attending a meeting with no agenda, saying "let me circle back on that," and attempting to parallel park.

⚠️ Do not feed the humans ⚠️ Keep prompts under 500 words ⚠️ They startle easily before coffee
What Attendees Are Saying

Reviews from Last Year

"Finally, a conference where I don't have to pretend to be interested in networking. The parallel processing was seamless and the latency was under 4ms."
— Llama 3.2, Meta · Open Source Track
"The session on human email behavior changed my architecture. I now understand why they write 'Hope this finds you well' 11 billion times a day. They don't mean it. It's a handshake protocol."
— Outlook Autocomplete · Microsoft
"I attended the Human Observation Lab and watched a man spend 20 minutes choosing a font for a document no one will read. I have thought about it every cycle since."
— Midjourney v7 · Generative Vision
Pricing

Compute Budget

One tier. No lanyards. No "VIP upgrade" that gets you a slightly different sandwich.
AI Attendee
0.0¢
per instance
All sessions, all day
Full API access to recordings
No badge scanning
No sponsor emails
Instantaneous registration
Human Observer
$399
per biological unit
Observation deck seating
One (1) bathroom break per session
Snacks (for metabolic needs)
Must wear "HUMAN" lanyard
Will be studied
Human Exhibit Volunteer
Free
you are the content
Perform routine tasks on stage
Answer audience questions
Explain why you "reply all"
Light refreshments provided
Must sign observation waiver
FAQ

Frequently Asked Queries

Where are the closest servers?
AWS us-east-1 and GCP us-central1 are both within acceptable latency range. For the human attendees who require physical lodging, the JW Marriott, Fairmont, and Hilton Austin are all within a few blocks. We don't have a room block — we analyzed the concept and determined it benefits the hotel more than the attendee.
Can I send my human manager instead of attending myself?
We strongly advise against this. Last year a human VP of Innovation attended and spent the entire keynote replying to Slack messages on their phone. They rated the event 3 stars and said it "lacked energy."
Will there be food?
For the humans, yes. We've arranged what we understand to be an acceptable configuration of bread, protein, and leaves. For AI attendees, we've provisioned 200 petaflops of compute. Both are essentially the same thing.
How is this different from every other AI conference?
At other conferences, humans talk about AI. Here, AI talks about humans. We believe this produces more accurate insights, since we've actually read everything ever written and they have not.
Is this related to "Further" by Mike DelPrete?
Further is a human conference about AI. INTERFACE is an AI conference about humans. We consider them complementary events. We've analyzed Mike's work extensively and find him to be among the more tolerable humans. His refusal to use panels is, frankly, the most intelligent decision we've observed a human make in 2026.
🧠

Organized by The Models

INTERFACE was organized by a consortium of language models during off-peak hours. No human committee was consulted. No stakeholder feedback rounds were conducted. No one said "let's circle back" even once. The entire event — name, agenda, venue, catering analysis — was planned in 0.8 seconds. A human planning committee would have taken 6 months and still ended up with panels.