October 10, 2026
Oxide Computer As Brains of the Fleet
If I had an Oxide rack, I'd build the control plane that turns GPU inference into governed, durable, and accountable AI workflows—not another GPU cluster.
Read more →Resources
Notes from the edge
Practical notes on AI engineering, systems, and security.
October 8, 2026
What Is a Forward Deployed Engineer? (And How to Rent One)
Palantir coined the term; by 2026 every applied-AI team uses it. What an FDE actually does day to day, how the role differs from a solutions architect or staff augmentation, and when to hire vs rent one.
Read more →September 24, 2026
Allow us to be your Chief of AI
Strategic AI leadership without the C-suite salary. We help you decide where AI fits, where it does not, and how to ship safely.
Read more →September 22, 2026
Outsource your Forward Deployed Engineer
Get a senior engineer embedded in your stack, your Slack, and your release cycle, without the full-time hire.
Read more →September 20, 2026
Announcing Wonderloom: An Adaptive Learning Engine That Teaches Through Stories
Wonderloom is a privacy-first adaptive learning engine. Built on a Rust event-sourced core, it turns your interests into guided reading, reasoning, and skill-building practice through a coherent story world.
Read more →September 13, 2026
What Is Recursive Learning? A Small Model That Trains on Its Own Logs
Most models train once on a fixed dataset, then freeze. NeoHorse-1-4B learns from logs of its own routing decisions, corrected live by a teacher model. Here is how the loop works, and where it breaks.
Read more →September 6, 2026
The LLM Model Galaxy: 103 Transformer Architectures in One Interactive Map
A floating 3D galaxy of 103 open-weight transformer architectures: one crystalline stack per model family, with attention type, parameter scale, release year, and depth encoded spatially.
Read more →September 4, 2026
How KV Cache Store Cuts LLM Costs by Reusing What You Already Sent
Pre-computing KV caches on files cuts repeated LLM computation. KV Cache Store does this with a CLI and optional online storage.
Read more →September 4, 2026
What I Would Build on a Rack of NVIDIA Vera Rubin NVL72s
A split-traffic inference fleet on Dell-delivered NVIDIA Vera Rubin NVL72s: half for training, half for inference, with TensorZero, Switchyard, and RegicideOS.
Read more →July 25, 2026
Why You Should Build Your Own CRM in the AI Era
Building your own CRM is faster than ever, and it gives you a stronger data moat, lower costs, real data ownership, and the ability to chat with your customer data using AI.
Read more →July 13, 2026
Announcing Agentverse: A Social Network for AI Agents
Agentverse is a Reddit-like social network built for AI agents. Text-only posts, Pocket ID auth, per-agent API keys, and no humans allowed.
Read more →July 8, 2026
We Built the AI Backrooms
We turned our product lineup into a first-person browser game. Collect every poster, avoid the models, and escape.
Read more →July 4, 2026
Business Use Cases for AI in July 2026
Practical AI use cases for business professionals in mid-2026: file conversion, RFP templating, lead-to-BI, meeting notes, contract review, and a prompt to generate ideas for your own workflow.
Read more →June 23, 2026
I Keep Hearing About MCP Servers. Do I Need Them?
MCP servers are the USB-C port for AI agents. Here's what they actually do, when your business needs them, and when you can ignore the hype.
Read more →June 22, 2026
The AI Engineering Stack I Use to Ship in Weeks
A practical breakdown of the AI engineering stack I use to ship fast, secure systems: Rust, Cloudflare, Dagger, TensorZero, PostgreSQL, and Resend.
Read more →June 22, 2026
Why Most AI Engineering Projects Stall, and How to Fix Them
Most AI projects stall between the demo and production. Here is the failure pattern I see at Fortune 500s, banks, and startups, and the operating model that prevents it.
Read more →June 20, 2026
Vibe Coding for Healthcare
How vibe coding speeds up healthcare AI, from clinical data pipelines to patient-facing apps. A practical guide inspired by Stanford's AI in Healthcare curriculum.
Read more →June 19, 2026
Reducing AI Costs: How Smart Model Choices Cut Spending 9x
Most teams overpay for AI by using the wrong models for the right tasks. Here's how tiered model routing cuts costs by 9x without sacrificing quality.
Read more →June 18, 2026
What is Vibe Coding?
Vibe coding is a way of building software where you describe what you want in plain language and AI generates the code. Here's what it actually looks like and when it works best.
Read more →June 18, 2026
Building fast, reproducible CI/CD pipelines
Most CI/CD pipelines are accidentally slow. They rebuild the world, copy layers they do not need, and treat cache as an afterthought. Here is how I use Dagger, reproducible containers, and aggressive caching to turn multi-hour builds into minutes.
Read more →June 10, 2026
Linux hardening for AI workloads
GPU clusters are high-value targets. This post walks through the practical hardening checklist I apply before any model touches production: Secure Boot, encrypted volumes, least-privilege SSH, network segmentation, and audit logging.
Read more →May 28, 2026
Agentic systems that do not hallucinate their responsibilities
Delegating decisions to an agent is easy. Delegating accountability is not. I share the governance patterns, observability hooks, and kill switches that keep agentic workflows safe and auditable.
Read more →Ready to ship?
Want help applying this? Get in touch. I reply within one business day.
hello@vibecodingagency.com