Skip to content
View jaywavy616-arch's full-sized avatar

Block or report jaywavy616-arch

Block user

Prevent this user from interacting with your repositories and sending you notifications. Learn more about blocking users.

You must be logged in to block users.

Content in all repositories owned by your account will be closed.
Maximum 250 characters. Please don’t include any personal information such as legal names or email addresses. Markdown is supported. This note will only be visible to you.
Report abuse

Contact GitHub support about this user’s behavior. Learn more about reporting abuse.

Report abuse
jaywavy616-arch/README.md (
## Hey, I'm Ethan

I build software by directing AI coding agents. Self-taught.

---

### Verso

Free push-to-talk dictation for Windows. Runs on device, no account, no subscription. Your voice never has to leave your machine.

- **148 commits** in about two and a half weeks
- **285 tests**, green CI on a Windows runner
- **0.0% word error rate** on clean speech, 1.2% at 10dB of noise
- **339 to 770ms** from key release to pasted text
- Dual-engine transcription: cloud Whisper raced against local GPU inference, with voice-activity gating so silence never produces hallucinated text
- MIT licensed

**[github.com/jaywavy616-arch/verso](https://github.com/jaywavy616-arch/verso)**

---

### Other things I've built

**ASTRA** is a local AI assistant. Long-term memory, an agent runtime, MCP servers, multi-agent orchestration, and model routing across local and cloud LLMs. 912 tests. I ran a security audit on it and found three real vulnerabilities by exploiting them first: a denial of service on the user-input path, a credential exfiltration route through unrestricted file reads, and a CSRF hole that let any web page launch its coding agent. All three are fixed. Private for now.

**A gaming clip detector** that finds the best moments in a stream by combining audio peaks, Whisper transcripts, and live chat velocity. Most tools in this space only listen for loud audio, which misses the quiet moments chat actually reacts to. Private.

**Postmatch**, a Next.js and Postgres site with CSP and HSTS headers, bot protection, and rate limiting. Built and shipped the whole thing.

---

### How I work

I build by directing AI coding agents and verifying what comes back. The verifying is the harder half, and it's the part I've gotten good at.

Right before I published Verso I caught three false claims in my own commit messages, including a test count that had been wrong across five commits in a row. The cause was subtle: I'd been adding test *functions* to a running total of *collected* tests, which silently breaks under parametrize.

That's the work. Ship fast, then assume you're wrong until you've checked.

---

### Also

I run a gaming channel on the side. One video past 3.6 million views, so I know how to make technical things reach people as well as function.

Open to remote work, worldwide.

Pinned Loading

  1. verso verso Public

    Free push-to-talk dictation for Windows. Runs on device with no account and no subscription. Your voice never has to leave your machine.

    Python