Long-form notes on software and AI engineering.
2026
Image search using text classification
I built a simple image search engine by distilling images into text labels using gemini 3.5 flash. No image embeddings, just a labeling agent and hybrid search.
Agent Platform That Builds Itself
I'm sharing my foundational agent platform that builds itself.
Auto-Improving Software
Coding agents have changed how we build software. Now they're changing how we improve it. In this post I share an agent platform that coding agents build, run, and improve on their own.
Agentic Software Engineering
Building, running, and scaling agents as production services
Context Providers
A thin layer between agents and tools that fixes context pollution, scope collisions, and prompt bloat.
How to Build an Agent Platform
A unified agent platform you can run locally or on Railway for $20/mo. Runtime, evals, scheduler, interfaces, and a Claude Code loop that recursively improves your agents.
Dynamic Software
For fifty years, software has been static. Then 2024 happened. The control flow came alive and a new category of software was born.
Systems Engineering
The Key To Building Agentic Software That Works
Scaling Agentic Software
What is the simplest architecture for running a multi-agent system at scale?
Dash: The Data Agent Every Company Needs
A self-learning, multi-agent data system. Free, open-source, runs in your cloud, lives in Slack.
The Programming Language for Agentic Software
The agentic era requires new primitives, a new execution model, and new runtime contracts
Build Your Own Multi-Agent System
Build a production-grade multi-agent system. Run locally using docker and deploy to production on Railway.
Evals Don't Give You a Working Product
Two years and billions in funding later, 80% of AI projects still fail. Are evals the solution or the problem?
Learning Machines: Technical Design
How Learning Machines work under the hood.
Learning Machines: Why AI Memory Hasn't Been Solved (Yet)
Every AI memory system I've used is missing something.
2025
Memory: How Agents Learn
How to build agents that are not only capable, but learn and improve over time.
GPU Poor Continuous Learning with Gemini 3
How to make agents better without fine-tuning or retraining. System-level learning that actually works.
Self Learning Research Agent That Tracks Consensus Over Time
Build a self-learning research agent that captures consensus, explains what changed, and improves over time.
Self Improving Text2Sql Agent with Dynamic Context and Continuous Learning
Learn how to build a text-to-sql agent with dynamic context and poor man's continuous learning.
Agent Security 101
Stop sending your agent transactions to telemetry services. Give your Agents a database and keep your data private.
WTF are Agents?
Understanding Agents by mapping out how they work.
Agent Engineering 101
A practical guide to Agent Engineering: the intersection of software, systems and security engineering.
Agentic Culture
An open-source experiment in collective memory and in-context cultural accumulation for multi-agent systems.
Becoming AI-first
Lessons from 100s of conversations on AI products and how teams are adopting AI.
Introducing Agno
The multi-agent framework, runtime, and UI built for speed.