---
title: "marcuschen (@marcuscodes) | Hashnode"
description: "marcuschen (@marcuscodes) writes on Hashnode. Read their latest articles and follow their work."
url: "https://hashnode.com/@marcuscodes"
---

@marcuscodes

# marcuschen

@marcuscodes · New York, USA · Joined May 2026

### [Three weeks before the enterprise contract, the voice agent wasn't operator-ready](https://voicelatency.hashnode.dev/three-weeks-before-the-enterprise-contract-the-voice-agent-wasn-t-operator-ready?utm_source=hashnode&utm_medium=feed)

Look. We had 99.2% uptime in staging. We had eval coverage on 1,400 test turns. We had latency under 280ms first-token. We were not operator-ready. I know this because the enterprise pilot started on

7 min read · [voicelatency.hashnode.dev](https://voicelatency.hashnode.dev/three-weeks-before-the-enterprise-contract-the-voice-agent-wasn-t-operator-ready?utm_source=hashnode&utm_medium=feed)

[Aug 19](https://hashnode.com/posts/three-weeks-before-the-enterprise-contract-the-voice-agent-wasn-t-operator-ready/6a85372452f5fb22ea3241a0)

### [Your ASR confidence score is a number you can act on. We were throwing it away.](https://voicelatency.hashnode.dev/your-asr-confidence-score-is-a-number-you-can-act-on-we-were-throwing-it-away?utm_source=hashnode&utm_medium=feed)

Voice agents treat the transcript as ground truth. The speech recognizer often tells you it is not sure, and we were ignoring it. TL;DR: When I went back through a month of "the agent did the wrong th

3 min read · [voicelatency.hashnode.dev](https://voicelatency.hashnode.dev/your-asr-confidence-score-is-a-number-you-can-act-on-we-were-throwing-it-away?utm_source=hashnode&utm_medium=feed)

[Aug 18](https://hashnode.com/posts/your-asr-confidence-score-is-a-number-you-can-act-on-we-were-throwing-it-away/6a83f5a7bbd6120da107f1fc)

### [The demo was flawless. The first real call had three people talking.](https://voicelatency.hashnode.dev/the-demo-was-flawless-the-first-real-call-had-three-people-talking?utm_source=hashnode&utm_medium=feed)

A scripted voice-agent demo works with one clean speaker. The first production call had crosstalk, and turn-taking fell apart. This is how I fixed it. The demo went perfectly. It always does. One pers

8 min read · [voicelatency.hashnode.dev](https://voicelatency.hashnode.dev/the-demo-was-flawless-the-first-real-call-had-three-people-talking?utm_source=hashnode&utm_medium=feed)

[Aug 13](https://hashnode.com/posts/the-demo-was-flawless-the-first-real-call-had-three-people-talking/6a7e46e5aa4a7908b9237cc2)

### [The call failed on turn nine. My eval gave me one number for the whole call.](https://voicelatency.hashnode.dev/the-call-failed-on-turn-nine-my-eval-gave-me-one-number-for-the-whole-call?utm_source=hashnode&utm_medium=feed)

The transcript was fourteen turns long and the score was 0.62. That is the entire output. One float, one call, and a rubric that said something like "did the agent resolve the customer's issue." It di

13 min read · [voicelatency.hashnode.dev](https://voicelatency.hashnode.dev/the-call-failed-on-turn-nine-my-eval-gave-me-one-number-for-the-whole-call?utm_source=hashnode&utm_medium=feed)

[Aug 13](https://hashnode.com/posts/the-call-failed-on-turn-nine-my-eval-gave-me-one-number-for-the-whole-call/6a7e3ff373398ffca6828656)

### [Two weeks before launch, every turn was green and the call still died](https://voicelatency.hashnode.dev/two-weeks-before-launch-every-turn-was-green-and-the-call-still-died?utm_source=hashnode&utm_medium=feed)

The dashboard was a wall of green. Word error rate under 5 percent. Intent classification at 94 percent on our eval set. Response appropriateness, graded by a rubric we trusted, sitting comfortably in

8 min read · [voicelatency.hashnode.dev](https://voicelatency.hashnode.dev/two-weeks-before-launch-every-turn-was-green-and-the-call-still-died?utm_source=hashnode&utm_medium=feed)

[Aug 10](https://hashnode.com/posts/two-weeks-before-launch-every-turn-was-green-and-the-call-still-died/6a7a5097209edc1ae5ee7c4f)
