Nathan Brooks

Building AI tools and software. Off-screen: fixing my bike, trying a new bread recipe, or losing track of time in a used bookstore.

Joined on Sep 7, 2026

Nathan Brooks

Sep 18

Building a Chat-to-Video App Without Coupling It to One Model Provider

#ai #agents

6 min read

Nathan Brooks

Sep 18

Long-Running Agent Prompts Need Contracts, Not More Instructions

#ai #agents

8 min read

Nathan Brooks

Sep 17

Kimi K3 Deployment: I’d Measure Accepted-Task Cost Before Renting GPUs

#ai #agents #opensource

7 min read

Nathan Brooks

Sep 17

Cutting Agent Token Costs Without Breaking Task Quality

#ai #agents

5 min read

Nathan Brooks

Sep 17

Building a Multi-Model Media Queue in n8n

#ai #agents

6 min read

Nathan Brooks

Sep 17

GPT-6 Sol: What the Leaks Tell Me—and What I Would Not Build Around

#ai #agents

6 min read

Nathan Brooks

Sep 16

Agents Need Feedback Loops. World Models Add a Way to Look Ahead.

#ai #agents

8 min read

Nathan Brooks

Sep 16

GLM-6.0 Is a Feedback-System Roadmap, Not a Model Spec

#ai #agents #opensource

7 min read

Nathan Brooks

Sep 15

One OpenAI-Compatible Endpoint for Multiple AI Models

#ai #agents #opensource

8 min read

Nathan Brooks

Sep 14

Claude API Access Without an Anthropic Account: Five Practical Routes

#ai #agents

9 min read

Nathan Brooks

Sep 10

GPT-6 Astra Costs: I’d Budget for Finished Tasks, Not Just Tokens

#ai #agents #opensource

9 min read

Nathan Brooks

Sep 9

My Claude Code Setup for Hard Tasks with Fable 5.1

#ai #agents #opensource

7 min read

Nathan Brooks

Sep 8

DeepSeek V5 Isn't Here Yet. Your Eval Harness Should Be.

#deepseek #ai #webdev #programming

3 min read

DEV Community

Nathan Brooks

Badges

Writing Debut

Building a Chat-to-Video App Without Coupling It to One Model Provider

Long-Running Agent Prompts Need Contracts, Not More Instructions

Kimi K3 Deployment: I’d Measure Accepted-Task Cost Before Renting GPUs

Cutting Agent Token Costs Without Breaking Task Quality

Building a Multi-Model Media Queue in n8n

GPT-6 Sol: What the Leaks Tell Me—and What I Would Not Build Around

Agents Need Feedback Loops. World Models Add a Way to Look Ahead.

GLM-6.0 Is a Feedback-System Roadmap, Not a Model Spec

One OpenAI-Compatible Endpoint for Multiple AI Models

Claude API Access Without an Anthropic Account: Five Practical Routes

GPT-6 Astra Costs: I’d Budget for Finished Tasks, Not Just Tokens

My Claude Code Setup for Hard Tasks with Fable 5.1

DeepSeek V5 Isn't Here Yet. Your Eval Harness Should Be.