---
title: "David Hahn (@davidhahn) | Hashnode"
description: "10+ years in software engineering, with deep expertise in frontend. Now going deep on LLMs — streaming, RAG, tool use, and everything in between."
url: "https://hashnode.com/@davidhahn"
---

@davidhahn

# David Hahn

@davidhahn · Chicago, USA · Joined June 2026

10+ years in software engineering, with deep expertise in frontend. Now going deep on LLMs — streaming, RAG, tool use, and everything in between.

### [Cost & Latency Tracking — What the Token Counts Were Telling Me All Along](https://blog.davidhahn.co/cost-latency-tracking-what-the-token-counts-were-telling-me-all-along?utm_source=hashnode&utm_medium=feed)

Every module up to this point used the same streaming events without ever reading two fields that were there the entire time. message\_start and message\_delta carry token usage on every single stream —

6 min read · [blog.davidhahn.co](https://blog.davidhahn.co/cost-latency-tracking-what-the-token-counts-were-telling-me-all-along?utm_source=hashnode&utm_medium=feed)

[Jun 19](https://hashnode.com/posts/cost-latency-tracking-what-the-token-counts-were-telling-me-all-along/6a349018aaf238351becba9c)

### [Error Handling in LLM Systems — Three Categories, One Decision Tree](https://blog.davidhahn.co/error-handling-in-llm-systems-three-categories-one-decision-tree?utm_source=hashnode&utm_medium=feed)

Everything in the earlier modules assumed the happy path: the API responds, the model generates valid output, tools return results, the stream closes cleanly. In production, none of that is guaranteed

7 min read · [blog.davidhahn.co](https://blog.davidhahn.co/error-handling-in-llm-systems-three-categories-one-decision-tree?utm_source=hashnode&utm_medium=feed)

[Jun 17](https://hashnode.com/posts/error-handling-in-llm-systems-three-categories-one-decision-tree/6a3307d01f612fccda42afc9)

### [Topic Suggestion — Designing a Function That Knows What to Recommend Without Magic Numbers](https://blog.davidhahn.co/topic-suggestion-designing-a-function-that-knows-what-to-recommend-without-magic-numbers?utm_source=hashnode&utm_medium=feed)

Phase 4 gave the study system memory of individual problems — when each one is due for review based on SM-2. Phase 5 asks a different question: zoomed out across all topics, what should I actually be

8 min read · [blog.davidhahn.co](https://blog.davidhahn.co/topic-suggestion-designing-a-function-that-knows-what-to-recommend-without-magic-numbers?utm_source=hashnode&utm_medium=feed)

[Jun 17](https://hashnode.com/posts/topic-suggestion-designing-a-function-that-knows-what-to-recommend-without-magic-numbers/6a31f725aca3c5bc517d5ddf)

### [Streaming Structured Output — Incremental JSON Rendering Without a Parser](https://blog.davidhahn.co/streaming-structured-output-incremental-json-rendering-without-a-parser?utm_source=hashnode&utm_medium=feed)

04-structured-output solved the reliability problem: use forced tool use to guarantee valid JSON back from the model. But it waited for the complete response before doing anything with it. For a backe

5 min read · [blog.davidhahn.co](https://blog.davidhahn.co/streaming-structured-output-incremental-json-rendering-without-a-parser?utm_source=hashnode&utm_medium=feed)

[Jun 15](https://hashnode.com/posts/streaming-structured-output-incremental-json-rendering-without-a-parser/6a3081a377097ef51547482d)

### [Evals — Why a Bad Eval Is Worse Than No Eval](https://blog.davidhahn.co/evals-why-a-bad-eval-is-worse-than-no-eval?utm_source=hashnode&utm_medium=feed)

The first time I ran evals on my LLM playground, the agent scored lower than basic RAG — 7.5 vs 8.9. The agent is more capable: it has live tool data on top of retrieval, it can answer questions basic

6 min read · [blog.davidhahn.co](https://blog.davidhahn.co/evals-why-a-bad-eval-is-worse-than-no-eval?utm_source=hashnode&utm_medium=feed)

[1](https://hashnode.com/posts/evals-why-a-bad-eval-is-worse-than-no-eval/6a2c7a4051327aaa82812867) [Jun 12](https://hashnode.com/posts/evals-why-a-bad-eval-is-worse-than-no-eval/6a2c7a4051327aaa82812867)
