@techwithj

Your APM Looks Green. Users Are Still Abandoning Calls.

Apr 16 · 8 min read · Standard APM was designed for deterministic systems. Voice AI is not a deterministic system. Your uptime is 99.9%. Your error rate is flat. Your average latency is 420ms. And users are abandoning call

Voice Agent Went From 66% to 96% Accuracy. Here's the Exact Optimization Loop I Used.

Apr 14 · 5 min read · Most voice agent testing loops look like this: you dial in, order something, decide it felt okay, and ship. I did this for longer than I should have on a drive-thru voice agent I was building called "

TTS API Comparison 2026: Choosing the Right Text-to-Speech Provider for Production

Apr 10 · 12 min read · Picking a TTS provider used to be a five-minute decision. You listened to a demo, grabbed an API key, and shipped. That shortcut breaks in 2026, when voice agents handle live customer calls and TTS de

What I Learned Testing Voice Agents at Scale: From 50 Manual Calls to 10,000 Automated Scenarios

Apr 7 · 10 min read · If you're building voice agents on Vapi, Retell, or any similar platform, you've done this: make a prompt change, spin up a test call, listen to the whole conversation, take notes, repeat. Maybe you p

LiteLLM Got Backdoored: Full Technical Breakdown, Incident Response, and Why Self-Hosted LLM Proxies Are a Liability

Mar 26 · 11 min read · On March 24, 2026, two malicious versions of LiteLLM landed on PyPI. Versions 1.82.7 and 1.82.8, published by a threat actor called TeamPCP, contained a three-stage payload that harvested SSH keys, cl