> field.init agents=1,600 humans=1
A practice by Henrique Jr. Sanchez
We taught everyone to press the buttons. Nobody taught them which problems the machine is for.
is the practice of working with AI deliberately: knowing which stance you are in, what you must supply for it to hold, how much of the work you are watching, and how big an ask you are making. It is for people who already use AI every day and still get burned, and for the teams that have to show they did something about it.
§1 four_axes
The Stances, the Stack, the Specifying guide with the AI @ Workplace curriculum, and Felix Rieseberg's ambition types are not separate things. Together they map four axes of one space. is the skill of moving through it on purpose, and re-entering at the right point when the work goes wrong.
What am I expecting from this exchange?
Oracle, Genie, Tutor, Mirror, Adversary, Librarian, Interpreter, Simulator, Scribe, Executor, Muse, Miner, Therapist, Translator.
See all fourteen stances →What must I supply for that expectation to hold?
Prompt, Context, Intent, Judgment, Coherence; plus two meta-functions, Evaluation and Harness.
Walk the stack →How much of the work do I watch?
Assistive (you type, AI suggests), Directed (you describe, AI types, you review), Autonomous (you assign, AI executes, you review), Continuous (AI runs as a background service).
Autonomy and spec rigor →How big is the ask?
Speed, Altitude, Depth, Range.
The four ambition types →§2 stance_selector
Every exchange with AI carries an expectation you may never have said out loud. Pick any of the fourteen and watch hundreds of tokens behave as that contract: where they go, what they ask of you, and where the work quietly fails.
Pointer or touch: you are the warm point.
stance Oracle
expects Truth to a hard question
load JUDGMENT
fail Confident plausible prose taken as fact
guard Ask for the uncertainty before the answer. Treat the answer as a claim, not a verdict.
§3 matrix.circuit
A stance is the contract you think you have with the AI. A Stack layer is what you must actually supply for that contract to hold. Each stance leans on one layer more than the others; that is where its failure lives, and where the guardrail belongs. Hover, tap or tab to a stance and follow the current.The full table is on the matrix page.
stance
layer
Oracle —load→ Judgment
§4 start_here
--role learner
Start with the AI @ Workplace video series (recorded talks with slides) at the level that matches your experience, then take the five-minute self-diagnostic to find your default stance.
--role team-lead
The matrix tells you which layers your shared setup must make strong, given the stances your team uses most. The assessment applies it to your real workflows.
--role builder
The Stack is the architecture. The Agent Audit Protocol in the repository is written for your agent to run against your own system.
The AI-Human Engineering Stack →Agent Audit Protocol (GitHub) →
§5 services
A fixed-fee diagnostic for a small team: where your AI work breaks, which layer to fix first, a 90-day plan, and an evidence pack of your AI-literacy measures.
from $3,500 · about two weeks
Remote, half or full day, mapped to your team's own workflows.
from $1,200 per session
Everything on this site and in the repo is free. You pay for my time.
Details§6 latest_essay
“True computer literacy is not just knowing how to make use of computers and computational ideas. It is knowing when it is appropriate to do so.”
Swap “computers” for “AI” and you have the problem this site is about.
Read the essay