---
title: "Sravan Vidiyala (@sravanvidiyala) | Hashnode"
description: "Engineering LLM systems that hold in production. Read the latest articles by Sravan Vidiyala on Hashnode."
url: "https://hashnode.com/@sravanvidiyala"
---

@sravanvidiyala

# Sravan Vidiyala

@sravanvidiyala · Sydney, Australia · Joined June 2026

Engineering LLM systems that hold in production.

### [How subagent sessions cascaded into a three-and-a-half-hour, 3,486-firing wiki-update pileup](https://blog.sravanvidiyala.com/how-subagent-sessions-cascaded-into-a-three-and-a-half-hour-3-486-firing-wiki-update-pileup?utm_source=hashnode&utm_medium=feed)

Every time I close a Claude Code session, a hook fires. It reads the session's transcript and decides whether anything in it is worth keeping. If so, it spawns a headless Claude process that distils i

10 min read · [blog.sravanvidiyala.com](https://blog.sravanvidiyala.com/how-subagent-sessions-cascaded-into-a-three-and-a-half-hour-3-486-firing-wiki-update-pileup?utm_source=hashnode&utm_medium=feed)

[Nov 3](https://hashnode.com/posts/how-subagent-sessions-cascaded-into-a-three-and-a-half-hour-3-486-firing-wiki-update-pileup/6ae9cb26e139425bbc5fe281)

### [My test suite was green. My detector was wrong.](https://blog.sravanvidiyala.com/my-test-suite-was-green-my-detector-was-wrong?utm_source=hashnode&utm_medium=feed)

I built a scanner to catch poisoned MCP tool descriptions, pointed it at 13,474 real ones, and watched it flag a tool called prompt\_injection\_scan as prompt injection.

9 min read · [blog.sravanvidiyala.com](https://blog.sravanvidiyala.com/my-test-suite-was-green-my-detector-was-wrong?utm_source=hashnode&utm_medium=feed)

[3](https://hashnode.com/posts/my-test-suite-was-green-my-detector-was-wrong/6a796091a1f0931c50030008) [Aug 10](https://hashnode.com/posts/my-test-suite-was-green-my-detector-was-wrong/6a796091a1f0931c50030008)

### [Isolate by dependency, not by load](https://blog.sravanvidiyala.com/isolate-by-dependency-not-by-load?utm_source=hashnode&utm_medium=feed)

Standard load-testing phases by traffic volume. For an agent with third-party upstreams and multi-call turns, attribution is the harder problem.

7 min read · [blog.sravanvidiyala.com](https://blog.sravanvidiyala.com/isolate-by-dependency-not-by-load?utm_source=hashnode&utm_medium=feed)

[Aug 9](https://hashnode.com/posts/isolate-by-dependency-not-by-load/6a7878e8bfa472cc827bd78b)

### [You can't load-test an LLM agent with a dumb mock](https://blog.sravanvidiyala.com/you-can-t-load-test-an-llm-agent-with-a-dumb-mock?utm_source=hashnode&utm_medium=feed)

Mocking the model lets you measure everything else. The model was the only thing that mattered.

5 min read · [blog.sravanvidiyala.com](https://blog.sravanvidiyala.com/you-can-t-load-test-an-llm-agent-with-a-dumb-mock?utm_source=hashnode&utm_medium=feed)

[Jun 4](https://hashnode.com/posts/you-can-t-load-test-an-llm-agent-with-a-dumb-mock/6a215060312f863e8c1be7fe)
