Articles
These articles collect product retrospectives, architecture research, and Agent product thinking beyond the project case study.
14 articles
Jun 11, 2026
After a month of AI-assisted Agent product development, the real problems were not code generation but product loops, state models, recoverability, workflow gates, and demo paths.
Jun 11, 2026
Agent product pages are not static information displays. Task semantics, state models, API contracts, event reducers, and recoverability should come before UI.
Jun 09, 2026
Default multi-agent parallelism is losing its appeal. The valuable direction is controlled parallelism built on task isolation, context governance, permission control, and result verification.
Jun 08, 2026
AI increases software output speed and also amplifies complexity. Mature product judgment treats complexity as a budget tied to visible user value.
Jun 08, 2026
As AI accelerates code, documentation, design, and execution, the product bottleneck moves from who can do the work to who can define the problem, manage complexity, review results, and own delivery.
Jun 08, 2026
Using Codex and Claude Agent Skills to examine how progressive disclosure packages organizational experience, workflows, scripts, and acceptance criteria as reusable context assets.
Jun 08, 2026
Using OpenHands Action / Observation as a baseline to compare ReAct, LangGraph, AutoGen, Claude Code, smolagents, and SWE-agent across abstraction boundaries and product tradeoffs.
Jun 08, 2026
Reframing AI cognitive compensation as Plato's inspiration plane: helping users discover what AI can do, describe tasks, create task packages, and turn success into playbooks.
Jun 08, 2026
From Agent Loop, AGENTS.md, Skills, Sandbox, Approvals, Hooks, and Compaction, this article examines how code agents govern what the model should see each round.
Jun 08, 2026
A system view of OpenHands from SDK, Agent Server, App Server, Sandbox, EventStream, Condenser, and storage layers, with reusable product architecture tradeoffs.
Jun 07, 2026
The final step from Task Context to LLM input must balance context governance, tool-call ordering, prompt cache hit rate, prefill latency, and audit boundaries.
Jun 07, 2026
Mainstream Agent products have mostly achieved multi-model access. Competition is shifting from provider count to file layers, capabilities, citations, usage, and tool-calling governance.
Jun 07, 2026
A look at Hermes Prompt Assembly, Context Engine, Context Compressor, Memory Provider, and Session Search as Agent context governance evolves into an auditable system capability.
Jun 07, 2026
Reframing Execution Agent context governance from prompt compression into a structured task-state, permission, evidence, and auditable decision contract.