I've been working on something for the past 3 months that I'm finally ready to share.
It started with a frustrating moment: I was coding at my desk when my AI agent hit a blocker. I wanted to walk away. I couldn't. I sat there for 45 minutes, watching a progress bar I couldn't leave.
This happens to developers constantly. AI coding tools made us faster, but they trapped us at our desks.
So I decided to build Cipher — AR glasses for supervising AI coding agents from anywhere on Earth.
Here's the problem:
Remote AI supervision exists (Claude Code, Cursor mobile), but the UX is broken. Squinting at code on a phone screen, typing corrections on a tiny keyboard, zero visual context. The capability is real. The experience is punishment.
Cipher fixes the interface:
- AR glasses — see code in context, not raw text on a tiny display
- Voice control — speak instructions, no keyboard friction
- Multi-agent team — 7 specialist AI agents (architecture, security, testing, DevOps, data, frontend)
- Hands-free — supervise while walking, same capability, 10× better experience
What's working today:
✅ 7 AI agents orchestrated in real-time (CrewAI + Groq models)
✅ Sub-1.5 second latency globally over internet architecture
✅ Real coding tasks completed successfully end-to-end
✅ I use it daily on my own projects
The plan:
- Phone app beta (next few months) — validate with waitlist users
- AR glasses prototype — build hardware in parallel
- Kickstarter launch — prove demand, deliver to early adopters
I'm 17.
I'm a full-stack developer.
I've been coding since 14.
I had an ML engineering internship at 15.
I AM the target customer — I built this because I needed it.
This is pre-seed stage.
I'm actively recruiting angel investors and co-founder for operations/legal/Funding (Since I'm 17, I need a rockstar partner to handle business entities, corporate structuring, and legal scaling while I run the technical engine)..
Building in public.
If you're a developer who's ever felt stuck babysitting AI Coding agents, this is for you.
👇 Join 12+ developers on the early-access waitlist. Link in the comments below!
Feedback, questions, criticism — all welcome.
Feedback, questions, criticism — all welcome.
If you're a developer who's ever felt stuck babysitting AI Coding agents, this is for you.
👇 Join 12+ developers on the early-access waitlist. Link in the comments below!
Feedback, questions, criticism — all welcome.
Top comments (3)
The reframe that saved me here: the goal was never to watch the agent, it's to make walking away from it safe. Two things got me there. Explicit stop conditions written into the instructions so it halts and asks at ambiguity instead of guessing and barreling on. And checkpoints, so a failure pauses the run instead of compounding quietly into 45 minutes of confidently wrong work. Once a run can stop itself cleanly, leaving the desk stops being a gamble. The progress bar was never the real problem, the silent compounding was.
@asdesbuilds Thank you — this is excellent feedback.
You nailed the core problem: “Make walking away safe” — not just possible.
That’s exactly what we’re optimizing for in Cipher.
Here’s how we’re thinking about it:
Explicit stop conditions → Every agent will have clear “halt & escalate” rules built in.
When it hits ambiguity, uncertainty, or high-risk decisions, it stops and presents options instead of guessing.
Checkpoints → We’re planning structured checkpoints after major steps (e.g., after writing a function, after refactoring a module).
The agent summarizes what it did, shows diffs, and waits for approval before continuing.
Clear communication → No silent compounding errors. Agents will explain the situation plainly + give 2-3 concrete choices.
This is one of the hardest (and most important) parts we’re working on right now.
Would love to hear more — how did you implement the checkpoints and stop conditions in your setup? Any patterns that worked especially well?
Appreciate the sharp insight.
If you think you'll find Cipher useful, join the waitlist : tally.so/r/rjGk9R
Some comments may only be visible to logged-in visitors. Sign in to view all comments.