Posts
Independent essays on AI-native work patterns, agent infrastructure, and what actually works when running agents as real tools.
-
When AI Makes Proof Cheaper Than Persuasion
Updated:AI makes working evidence cheap enough to replace part of organizational persuasion: build the prototype, let people test it, then decide.
-
The Right LLM for Each Step in an Agent Pipeline
Updated:Match model tier to cognitive load. A fast, single-call path with pre-fetched context cuts a 25-30 second two-call response to under 10 seconds.
-
Git-Native Agent Skills Need a Release Lifecycle
Updated:Tessl adds versioned publishing, immutable installs, and deliberate updates to repo-based skills. CI review exists, but teams must require it.
-
Name AI Skills for the Practice, Not the Output
Updated:An AI skill's name frames the relationship before the first run: output nouns create consumers; practice names invite users to internalize the workflow.
-
The Person Is the Constant: Context Selects the AI Voice Mode
Updated:Keep identity and quality standards constant, then select register from recipient, channel, and intent. One averaged profile blurs every context.
-
An AI Skill Should Leave You Better Without It
Updated:A well-designed AI skill exposes its reasoning sequence so repeated use builds practitioner judgment, not dependence on another output generator.
-
One Coding Agent, Three Bedrock APIs: Wiring OpenCode to Claude, GPT, Grok, and Kimi
A tested OpenCode configuration for running Claude, OpenAI GPT, xAI Grok, and Moonshot Kimi through their native Amazon Bedrock Runtime APIs.
-
Jev Does Not Write. That Is the Point.
TypeSafe's Jev returns a typed answer and a probability, not a paragraph. That is worth putting first on a clean enum, and worth nothing on a dirty one.