Robot or Exoskeleton?
Before building with AI, decide whether the work calls for autonomous execution or a human-augmenting exoskeleton.
ReadThinking
Essays, research, and analysis
On AI infrastructure, model behavior, and the economics of language models in regulated industries.
Lens
Eight Words
No essays match these filters yet.
-
-
Measure first; change weights last
Sequence AI investment through Gateway, policy, Knowledge, and routing before Training spend; measurement precedes weight updates.
Read -
Design AI surfaces around each role's workflow
Design AI interfaces by role—customer, operator, billing, executive—with one orchestration backbone and surfaces that match how each job actually runs.
Read -
OrchestrationStrategy June 9, 2026
Put human checkpoints in orchestration design
Design AI orchestration with explicit human approve, edit, and takeover points instead of optimizing for vanity deflection.
Read -
Sign the tools perimeter before systems credentials
Approve a Tools perimeter—systems, actions, human gates—before agent demos turn into production blast radius.
Read -
Put ground truth in the knowledge layer
Executive framework for governed Knowledge—corpora, citations, refresh—instead of fine-tuning on PDFs that expire before the weights ship.
Read -
System MessageStrategy June 6, 2026
Treat system messages as executable policy
Govern AI system messages like policy with owners, reviewers, prohibitions, and version dates instead of disposable prompt experiments in a sprint branch.
Read -
Own the gateway before you pick a model
Fund the Gateway before model contracts: routing, logging, data boundaries, spend caps — before negotiating another frontier model.
Read -
Match model tier to workflow ambiguity
Route by ambiguity and consequence so structured work uses cheaper model tiers and frontier spend goes where language is genuinely open-ended.
Read -
ModelImplementation May 19, 2026
LLM pricing runs on the alphabet
LLM tokenizers charge different languages different prices for the same meaning. Here are the real multipliers and what production teams can do about them.
Read
Bring us the AI decision you're trying to make.
We'll help you see the parts clearly.
Start a conversation