Posts
Independent essays on AI-native work patterns, agent infrastructure, and what actually works when running agents as real tools.
-
Scope the 1Password Vault, Then Scope the Token
Updated:A purpose-specific 1Password vault and read-only service account bound an always-on agent box to five secrets and blocked writes as configured.
-
1Password op run Makes Vault-Backed Credentials Tool-Transparent
Updated:At process launch, op run resolves 1Password references into ordinary environment variables, so CLIs and MCP servers need no vault-specific code.
-
AgentCore Code Interpreter: When the Agent Has to Actually Run the Math
Updated:AgentCore Code Interpreter ran a Pareto frontier over 528 models and matched the known-good result; tampering collapsed 13 results to one.
-
AgentCore Memory: What an Agent Remembers When the Session Is Gone
Updated:AgentCore Memory recalled preferences across microVM sessions and isolated three tenants, but asynchronous extraction can briefly return stale values.
-
AgentCore Runtime: Where an Agent Actually Runs
Updated:AgentCore Runtime hosts each agent session in a managed microVM behind an HTTP server contract. Seven test sessions ran with zero errors.
-
A Never-Say List Is the Fingerprint of an AI Voice
Updated:Writing samples teach cadence. Corrections and mode-specific never-say lists encode the rejected choices that make an AI voice distinctive.
-
Browser Use vs. AgentCore Browser: Two Managed Browsers, and Which Layer Each One Wins
Updated:AgentCore Browser supplies governed Chromium; Browser Use adds a natural-language agent and cached reruns. Eleven experiments expose the tradeoffs.
-
Docs-as-Code as a Knowledge Distribution Stack
Updated:Markdown, Git, llms.txt, and MCP create human and agent read paths from one source, while AI-mediated pull requests remove Git fluency from writing.