Chert × HackUTA
Hack UTA
Best Use of FaceTime
Chert puts AI agents in iMessage and FaceTime. Build one that’s more useful when it can see.
Start building
- Copy the prompt below.
- Paste it into Claude Code, Cursor or Codex. No forms, about 5 minutes.
- That’s it. You’re on Chert with a free Chert line for 48 hours.
$4001st place
Top 3 teams get guaranteed interviewsfor our FDE and SWE intern roles. We’re currently expanding our team and actively hiring.
Need ideas? Find Gary at the Chert table.
iMessage
Hey HackUTA! I’m Chert. Text me anything about the FaceTime track.
Prizes, rules, what to bring, an idea to build, or how to reach the team.
What Chert does
We put AI agents in iMessage and FaceTime.
We treat FaceTime as a channel for realtime video. Your agent sees the caller’s camera live, sends video back, and can call people when something comes up.
- Sees the caller’s video live
- Sends video back
- Group FaceTime calls
- Places and answers calls
- Texts over iMessage
It can be a character you talk to, or no character at all. Build an agent that gets more useful the moment it can see.
Backed by Y Combinator (P26), OpenAI, Z Fellows and angels.
Uh, which cable?
“The blue cable, second from the left.”
A demo we love
ClawdTime: your coding agent FaceTimes you.
You step away from your laptop. Claude Code or Codex hits a decision. Your phone rings. You answer, say “use the blue heading,” and the agent keeps working on your computer.
It’s your computer, from anywhere, over a FaceTime call. Built on Chert. This is the kind of build we want to see.
See ClawdTimeGive your agent eyes.
Use the blue heading.
Done. Want me to push it?
To qualify
- Build on Chert’s FaceTime APIiMessage is optional, but encouraged.
- Bring an iPhoneAt least one per team.
- Call during your demoYour agent makes or takes a FaceTime call.
When you submit on Devpost, select “Best Use of FaceTime (Chert)”.
How it’s judged
Live demo, judged in person by Gary Gao, CEO of Chert.Show it working on a real FaceTime call.
Template ideas
Your own idea wins.The best projects are ones we haven’t seen. Use these only as a starting point, and if you build on one, put a real spin on it. “Fresh use of FaceTime video” counts double in judging.
Agents
Your background agent calls you
A long-running coding or ops agent hits a decision, places a FaceTime call, shows you the diff, logs or chart as its video, takes your spoken answer, and resumes.
Under the hood
- Agent hook fires on blocker
- Chert places the call
- Render terminal or diff to video frames
- Speech to text → tool call → resume
Your spin: What else runs unattended? Training runs, deploys, scrapers, trading sims.
Hardware
Breadboard bring-up copilot
Point your camera at a breadboard or PCB. The agent reads the parts and wire colors, checks them against your schematic, and sends back an annotated frame of what’s wrong.
Under the hood
- Sample frames at 1–2 fps
- Vision model + part detection
- Diff against netlist or KiCad export
- Draw boxes, stream the frame back
Your spin: Swap the schematic for a 3D-printer, a robot arm or a car engine bay.
Group calls
Group-call game master
Everyone joins one group FaceTime. The agent watches the table, tracks the board state, rules on disputes, and shows a live scoreboard as its video.
Under the hood
- Group call with every player
- Board-state model per frame
- Game state machine + rules engine
- Render scoreboard as outgoing video
Your spin: Chess, poker, a D&D table, a sports drill, a study-group quiz.
Procedures
Step-by-step verifier
The agent turns a procedure into a step graph and won’t advance until it sees each step done correctly on camera. Skip one and it calls you back.
Under the hood
- Parse the doc into a step graph
- Per-step visual check with a confidence bar
- Low confidence → ask for a better angle
- Log proof frames per step
Your spin: Lab protocols, IKEA builds, medication setup, bike repair.
Memory
Visual memory you can text
During calls, the agent indexes what it sees. Later you text “where did I leave my keys?” and it answers with when and where it last saw them.
Under the hood
- Embed sampled frames
- Store with time and place in a vector DB
- iMessage question → retrieval
- Answer by text, or call to show it
Your spin: Inventory for a lab or a store shelf, a home move, a closet.
Accessibility
Walking guide for low vision
The agent rides along on a call while you walk, reads signs and crossing signals, flags obstacles, and only speaks when it matters. When it’s unsure, it adds a sighted friend to the call.
Under the hood
- Stream the rear camera on a call
- Detect obstacles, text and signals
- Priority queue so it talks only when needed
- Low confidence → add a friend to a group call
Your spin: Grocery aisles, airports, campus buildings, reading a medicine label.
Questions
The usual ones. Anything else, text Chert on the phone up top.
How do we get set up?
Copy the setup prompt into Claude Code, Cursor or Codex. It signs you up and connects your agent to Chert in about 5 minutes. No form needed.
Do we need to pay for anything?
No. Paste the setup prompt and your team gets a free Chert line for 48 hours.
Does my agent need a face or avatar?
No. It can be a character or no character at all. What matters is what it does with what it sees.
Can we use iMessage too?
Yes. A lot of strong builds text first, then call when they need to see something.
What do we need to bring?
At least one iPhone per team, plus the laptop you’re building on.
What do the winners get?
1st place gets $400. Top 3 teams get guaranteed interviews for Chert’s FDE (forward deployed engineer) and SWE intern roles. We’re currently expanding our team and actively hiring. Chert is backed by Y Combinator (P26), OpenAI, Z Fellows and angels.
Can the agent share or control my screen?
Not yet. Today the agent sees the caller’s camera, sends video back, and joins one-to-one or group FaceTime calls.
Give your agent eyes this weekend.
Paste one prompt into your coding agent. You’re on Chert in about 5 minutes.
Stuck? Ask in the Chert Discord or find Gary at the Chert table.