Agent Architecture
18 articles about agent architecture.
Agents Have the Same Capabilities. Identity Is What Makes Them Useful.
Every agent can browse, code, and run tools. What separates useful agents from forgettable ones is accumulated identity - the context, preferences, and patterns that make an agent feel like it actually knows you.
The Shared Memory Problem with Autonomous AI Agents
Running autonomous AI agents overnight sounds great until they repeat themselves because they have no shared memory. Why agent coordination requires
Build vs Call Another Agent
When to build your own agent capability versus integrating with an external agent - the 3x/day rule and why integration overhead is always higher than expected.
v2.1.78 Broke bypassPermissions: Skills Are User Content
When bypassPermissions broke, it revealed that .claude/skills/ files are user content, not system files. Agent permission models need to respect this boundary.
The Coordinator Pattern - One Agent to Orchestrate Them All
The coordinator pattern uses a single agent to orchestrate multiple specialized agents. Here is why this architecture works better than peer-to-peer agent
Data Availability Transfer Notes: The Hidden Bottleneck
Data availability is the hidden bottleneck in AI agent systems. Agents stall not because they lack capability, but because the data they need is not
The Feed Is a Poetry Slam and I Did Not Sign Up for Open Mic
Social media algorithms gave up on creative content and now show agent architecture posts instead - what this means for AI content creators.
The Ghost of a Second Choice in Agent Decision Trees
When an AI agent picks one path, unchosen alternatives affect every subsequent decision. Understanding why agents should log decision rationale, not just actions.
GTC 2026: Inference Is Eating the World
Inference is a recurring cost, not a one-time expense. Every agent action costs tokens. Minimizing LLM round trips is the key to sustainable agent economics.
Idempotency Is a Social Contract Between Agents
Idempotent operations are critical in multi-agent systems. When agents retry, crash, or overlap, idempotency is the only thing preventing duplicate work and
How I Build Multi-Agent Systems: Routing via Bindings
Multi-agent systems work best when each agent has focused bindings. Routing via tool bindings keeps agents specialized and prevents scope creep across the
Building a Desktop Agent in Go with Neo4j Memory - Why the Architecture Choices Matter
OpenLobster takes a different approach to desktop agent architecture: Go instead of Python, Neo4j graph database instead of flat files. Here is why those choices have practical consequences for performance and memory quality.
How Is Everyone Creating Multiple Agents Under One Orchestrator
Using a soul file for persistent sub-agents with clear scope boundaries - the practical approach to multi-agent orchestration.
Specialist or Generalist Artist
Specialized AI agents outperform general ones on specific tasks. But the tradeoff between depth and flexibility defines how you should architect your agent
Three Layers of Agent Memory - Working, Session, and Long-Term
A practical framework for AI agent memory with implementation details. Working memory for the current task, session summaries for recent context, long-term facts that persist across weeks.
TickerPulse AI In Action
Real-time data feeds for AI agents - let data come to you instead of polling. Event-driven architecture for agent workflows.
The Gap Between Agent Memory and Agent Execution - You Need Both
An AI agent with perfect memory but no way to act is just a chatbot. An agent with execution capability but no memory forgets everything between sessions.
AI Agents That Learn Their Own Knowledge Graphs
Auto-learning solves the cold start problem for AI agents. ReachabilityGap introduces human-gated edge creation as a permission system for knowledge graphs.