Building now
TomatoRTC: live infrastructure beyond the video call.
A programmable participant fabric for humans, agents, devices, browsers, and services operating in one coordinated real-time system.
Explore TomatoRTC ↗Founder + systems architectTechnology + creative practiceTaiwan / global
✦ NATHANIEL “NAT” CURRIER / TOMATORTC / NAT.IO
I build and lead complex AI, cloud, WebRTC, and real-time systems, connecting architecture, engineering execution, product strategy, and customer trust.
This is also where I explore food, health, music, relationships, creative practice, and the small systems that shape a meaningful life.
Current professional practice
I founded TomatoRTC to build real-time participant infrastructure for products where people, AI agents, devices, and services share live context. I also work across CTO, VP Engineering, Chief Architect, advisory, and founder-in-residence mandates where technical complexity has become a product, delivery, or customer-trust problem.
Building now
A programmable participant fabric for humans, agents, devices, browsers, and services operating in one coordinated real-time system.
Explore TomatoRTC ↗Practicing
Small health systems designed for metabolic stability and ordinary busy days.
The meal system ↗Learning
Consistency in music, writing, and technical work without flattening curiosity.
The practice ↗
Jul 2026 · 11 min read
The risk of AI-generated code is not simply that it can be wrong. It is that teams lose the learning conversation that turns questionable choices into better engineering judgment.
Read the field note ↗Question of the day · Today
Stay awhile if
Probably skip it if
Technology matters when it expands human agency. This is a working notebook for technology, life systems, creative practice, and big-picture views.
Read the manifesto ↗
Rate limiting looks like arithmetic in tutorials, but in production it allocates scarce capacity, encodes fairness assumptions, and shapes client behavior under stress.
Read ↗
RAG became the default way to ground LLMs on enterprise data, but that did not solve AI reliability. It exposed a harder reality: retrieval is infrastructure, and the real work is systems design, governance, and evaluation.
Read ↗
Shadow AI is not primarily a compliance failure. It is what happens when capability arrives before permission, and when demand outruns an organization's ability to govern what is already being used.
Read ↗
Some low-output days are not failures. They are recovery days. The difference between drift and compounding is whether you know how to protect your creative system when capacity drops.
Read ↗Pick one, run it this week, and document what changed. The goal is action, not perfection.
Verify that one API boundary behaves like an explicit contract under success and failure.
Reclaim deep-work time by removing friction from one recurring task sequence.
Protect momentum on a low-capacity day by shipping a reduced but intentional creative session.
Improve one project update so stakeholders can evaluate your capability without guesswork.

From the kitchen
All the tangy‑mustard bite of the Carolina classic, sweetened with allulose and a hint of blackstrap molasses for authentic depth but just 1 g net carb per serving.
Cook this ↗About this space