Dakota Codes
Dakota's developer blog - thoughts on software development, programming, and technology
Latest posts
Untitled
Untitled
Untitled
Untitled
// post about evaluations, KPS, metrics, fine-tuning
Untitled
Untitled
Untitled
macOS Daemons + Local AI: Your Mac Can Just Do Things
launchd has been sitting on your Mac this whole time, waiting to react to events for you. Add a local AI model and it gets kind of magical.
I redesigned dakotacodes in under 30 minutes with Claude Design and Claude Code
How I used Claude Design to shape a new visual system for my site, then handed that system to Claude Code to implement the UI shift across the actual codebase.
Open Models, Real Value: A Practical Guide for Enterprise Teams
A practical guide to open and local AI models for enterprise teams: what changed, what it costs, how to start, and how to build hybrid systems that actually work.
Visual Timers
A simple tool that makes timeboxing feel tangible. Here's why I like mine.
Sometimes, Production Is You
Running an always-on AI agent on $15 hardware. PicoClaw on a Raspberry Pi Zero 2W: what it is, what it teaches, and why personal infrastructure matters.
If There Isn’t a Model for That, Make Your Own
How to move from generic LLM behavior to domain-fit intelligence using Hugging Face, fine-tuning, and workflow-specific evaluations.
Designing Hybrid AI Systems: Local + Frontier Without the Drama
How to operationalize local and frontier models together with routing, observability, governance, and team-ready delivery patterns.
OpenClaw on a VPS: A Practical Security Learning Lab
Set up OpenClaw on a VPS while learning real-world secure networking: SSH hardening, firewalling, reverse proxying, least privilege, monitoring, and incident readiness.
Local Language Models: What Leaders and Builders Need to Know
A practical guide to local language models: definitions, tradeoffs, hands-on ways to try them, and patterns for building real applications on top.
Build How You Want
Agents are changing how software gets built, and individuals now have more leverage than ever. Build in the mode that fits your context, and do it intentionally.
How to Run C on an Apple Device (And Why You'd Want To)
I ported a tiny GPT to an Apple Watch in both Swift and C. The C version was 50x faster. Here's the bridging setup and why the gap exists.
The Three-Token Problem
Small perturbations, big shifts. What chaos theory and the three-body problem can tell us about prompt sensitivity, superposition, and the math we might need to begin mapping it.
The Knowledge Department
Announcing a research and systems project exploring how AI agents can safely create, maintain, and govern shared organizational knowledge.
Building RAGTime: A Complete RAG System That Runs Entirely in Your Browser
How I built a privacy-first RAG pipeline using WebLLM, Transformers.js, and IndexedDB.
Announcing ted-mosby: Your Open-Source Architectural Wiki Agent
ted-mosby is officially available on npm. Generate comprehensive, traceable architectural documentation for any codebase.
I love Macropads
A programmable keypad might sound like overkill. It's not.
Agent-Generated Architectural Wikis are really neat
How AI agents are solving the eternal problem of stale documentation by automatically generating comprehensive, traceable architectural wikis from your codebase.
Watch the Commits
An age-old tip for staying sharp with your tools, more relevant than ever.
A Few Ways to Train Your Own Models
A walkthrough of model training: what it actually is, the different approaches, and where to get the compute.
Fun Feature Breakdown: MiniTV
A floating YouTube player disguised as a tiny retro television. Because why not.
The Accidental Browser on Your Wrist
watchOS doesn't have Safari, so developers use an authentication API as a backdoor browser. It works, mostly.
Agent Skills Work for Humans Too
Agent Skills are the hot topic in AI-assisted development, but their real power is something simpler: documented workflows that anyone can follow.
Happy New Year 2026
Wishing everyone a wonderful 2026 filled with experimentation, learning, and building cool things
The Levers of Control in Claude Code
A deep dive into Memory, Slash Commands, Skills, Subagents, and Hooks—and when to pull which lever.
1.44 MB is Enough: Floppy Disks as Analog Computing
A technical exploration of how modern browser APIs and floppy disks enable a renaissance of physical, portable computing—combining web technologies with retro media for apps you truly own
Git Bisect: Binary Search for Bug Hunting
How to find the exact commit that introduced a bug using git bisect's powerful binary search algorithm
Claude's Artifacts: Another Tool for Rapid Prototyping
Claude's experimental artifacts feature joins tools like v0.dev and Lovable in the growing ecosystem of idea-to-app builders.
Claude Code: Your AI Partner for Any Text-Based Work
Discover how Claude Code goes far beyond programming to become an incredibly powerful interface for scientific research, writing, home management, and any text or image-based knowledge work.
Envisioning a Closed Agentic Loop for Software Development
An experiment in tooling that explores how development teams can benefit from AI with minimal friction while maintaining complete understanding and control over their codebase.
Git for Everyone: How Non-Programmers Can Benefit from Version Control
A no-knowledge-assumed introduction to Git that shows how version control can revolutionize any text-based work—from writing and research to project management and creative work.
Shipping Your First Feature With AI - Choose Your Own Agent!
An accessible, interactive guide for shipping your first feature using AI agents - no coding experience required. Choose your preferred AI assistant and follow along!
📎 It looks like you're writing user stories from API docs…
How I used AI to transform complex API documentation into actionable development tickets in minutes.
Getting Started with Cursor CLI: The Terminal-Lover's Path
A quick guide to using Cursor's terminal-based AI coding assistant — the perfect companion to the Claude Code tutorial.
Shipping your first real feature with Claude Code
An accessible, step-by-step guide for shipping your first feature using agentic AI — no coding experience required.
AI + iPhone = software development on the go
Exploring the possibilities of using an iPhone as a legitimate development machine with the help of AI tools like GitHub Copilot, and how this combination might shape the future of mobile development.
Back to the (Machine Learning) Future
A simple Gregg shorthand test shows that while the latest foundational models like GPT-5 are advancing quickly, traditional ML/DL techniques remain essential for specialized tasks.
The State of Coding Agent Models: August 2025
A practical guide to today’s coding agents: how agentic tooling works, how to pick the right model “brain,” and how to read benchmarks and leaderboards.
Zero to One
Post about building a custom MCP server to generate open graph preview images for my blog posts.
Hello World: Welcome to My New Blog
Welcome to my new React + Vite powered blog. Let me tell you about the journey from Astro to this modern setup.