- All
- Riteangle 11
- Operations 6
- Agent architecture 3
- Advertising 2
- Agentic architecture 2
- Context engineering 2
- Marketing 2
- Guardrails 1
Agent architecture — 3 posts
Five disciplines of building with models
Data, prompt, context, loop, harness. What each one actually engineers, what it costs, how slow it is to build, and the specific tools you need for each. Every layer assumes the one beneath it.
10 min read
From pipelines to loops
Two years of attending India's data and AI conferences. The data substrate barely moved. Four new floors were built on top of it, and the unit of engineering work shifted from the prompt to the loop.
8 min read
Match decisions without a model in the loop
riteangle runs LLMs at both edges of the matching pipeline and nothing in the middle. The decision itself is a weighted dot product and a min-cost max-flow solve. Here is the architecture, why the industry converged on it, and where we still violate it.
12 min read