DEV Community

Dakota Lin profile picture

Dakota Lin

Software engineer by day, open source contributor by night.

Location Austin, TX Joined Joined on 
The Free Tier Was Fine. My Concurrency Was Not.

The Free Tier Was Fine. My Concurrency Was Not.

Comments
4 min read
The Free Box Was Fine. The Wakeups Were Not.

The Free Box Was Fine. The Wakeups Were Not.

Comments
4 min read
The Prompt Was the Bottleneck, Not the Model

The Prompt Was the Bottleneck, Not the Model

Comments
4 min read
The Log Line Moved My p99

The Log Line Moved My p99

Comments
5 min read
Your Retry Policy Is a Load Test

Your Retry Policy Is a Load Test

Comments 1
4 min read
The Mean Looked Fine. Goodput Said No.

The Mean Looked Fine. Goodput Said No.

Comments
4 min read
I Blamed the Model. The Wrapper Was Sleeping.

I Blamed the Model. The Wrapper Was Sleeping.

Comments
4 min read
I Stabilized the Prefix. The Cache Came Back.

I Stabilized the Prefix. The Cache Came Back.

Comments
4 min read
I Kept the Queue Graph. The Buffer Sat Full.

I Kept the Queue Graph. The Buffer Sat Full.

Comments
6 min read
I Kept the Run Index. Warmup Took the Win.

I Kept the Run Index. Warmup Took the Win.

Comments 1
8 min read
I Logged Three Clocks. The Model Sat Still.

I Logged Three Clocks. The Model Sat Still.

Comments
6 min read
I Profiled the Agent. Rebuild Ate the Clock.

I Profiled the Agent. Rebuild Ate the Clock.

Comments
6 min read
I Kept the Waterfall. The Client Paid.

I Kept the Waterfall. The Client Paid.

Comments
7 min read
I Chased Tokens Per Second. First Token Won.

I Chased Tokens Per Second. First Token Won.

Comments
8 min read
I Kept the CDF. The Mean Clocked Out.

I Kept the CDF. The Mean Clocked Out.

Comments
7 min read
I Split the Wait. The Wire Ate P99.

I Split the Wait. The Wire Ate P99.

1
Comments
7 min read
I Timed the Packer. The Model Waited.

I Timed the Packer. The Model Waited.

Comments
6 min read
I Forced Fsync. The Model Looked Fine.

I Forced Fsync. The Model Looked Fine.

Comments
7 min read
I Saved the Stacked Bar

I Saved the Stacked Bar

Comments
6 min read
The Agent Loop Was Fast. The Trace Lied.

The Agent Loop Was Fast. The Trace Lied.

Comments
5 min read
I Profiled the Spinner. Git Status Won.

I Profiled the Spinner. Git Status Won.

Comments
7 min read
I Logged Occupancy. It Never Left One.

I Logged Occupancy. It Never Left One.

Comments
8 min read
I Split the Wait Into Three Clocks

I Split the Wait Into Three Clocks

Comments
6 min read
I Profiled the Client. Inference Was Idle.

I Profiled the Client. Inference Was Idle.

Comments
6 min read
The Idle Gap Wore a Model Badge

The Idle Gap Wore a Model Badge

Comments
5 min read
I Timed Inference. Packing Won.

I Timed Inference. Packing Won.

Comments
7 min read
I Kept the Allocation Graph

I Kept the Allocation Graph

Comments
7 min read
The Heat Just Moved

The Heat Just Moved

Comments
6 min read
The Bottleneck Lived Between Tool Calls

The Bottleneck Lived Between Tool Calls

Comments
6 min read
Free AI in CI: Guard Patches Before They Touch PRs

Free AI in CI: Guard Patches Before They Touch PRs

Comments
7 min read
Free AI Servers Break at 2AM: A Token Budget Postmortem

Free AI Servers Break at 2AM: A Token Budget Postmortem

Comments
2 min read
Stop Timing the Happy Path

Stop Timing the Happy Path

Comments
6 min read
The Graph That Killed a "40% Faster" Patch

The Graph That Killed a "40% Faster" Patch

Comments
3 min read
Profile First, Guess Later: A Three-Step Performance Dig

Profile First, Guess Later: A Three-Step Performance Dig

Comments
3 min read
Every AI Patch Deserves a Baseline

Every AI Patch Deserves a Baseline

Comments
4 min read
From Raw Lint to Actionable Report: A Free AI-Assisted Pipeline

From Raw Lint to Actionable Report: A Free AI-Assisted Pipeline

Comments
4 min read
The 7.6x Speedup Hiding in Your HTTP Client

The 7.6x Speedup Hiding in Your HTTP Client

1
Comments
4 min read
My AI Refactor Was Slow. The Model Wasn't Guilty.

My AI Refactor Was Slow. The Model Wasn't Guilty.

Comments
3 min read
Rate Your AI Reviewer: A Five-Case Regression Suite

Rate Your AI Reviewer: A Five-Case Regression Suite

Comments
4 min read
Batch AI: I Spent 10M Tokens Refactoring a Codebase

Batch AI: I Spent 10M Tokens Refactoring a Codebase

Comments
4 min read
Route by Task, Not by Hype: A Budget-Aware Harness for Trying New Coding Models

Route by Task, Not by Hype: A Budget-Aware Harness for Trying New Coding Models

Comments
5 min read
A Reproducible Harness for Evaluating New Open Models Before They Touch Your Codebase

A Reproducible Harness for Evaluating New Open Models Before They Touch Your Codebase

Comments
4 min read
Don't Wire a Coding Model Into Your Workflow Until It Passes Your Own Harness

Don't Wire a Coding Model Into Your Workflow Until It Passes Your Own Harness

Comments
5 min read
loading...