The irreducible epistemic atoms underlying the curriculum. 4,828 atoms across 8 types and 2 molecules
Before delegating a cognitive task to a tool, classify whether the delegation is appropriate (tool does it better and verification is easy), convenient (saves time but you could do it), or critical (you cannot perform it without the tool), because critical delegations create capability gaps when tools fail.
For critical tool delegations where failure would compromise your effectiveness, maintain periodic unassisted performance of the delegated capability to prevent atrophy of the biological skill, treating the practice as architectural redundancy rather than inefficiency.
Design tool-use policies that define which tools you allow for which tasks, under what conditions, and with what fallback behavior when tools fail, treating this as governance of your extended cognitive architecture rather than as tool preferences.
Verify habit automaticity by checking whether the behavior fires from context cues with minimal conscious effort rather than checking execution frequency, because consistency maintained through willpower is not delegation—true automaticity means the cue triggers the routine without deliberation.
Protect the habit formation transfer period (first 2-4 weeks) with environmental supports that make cues more salient and routines easier to initiate, treating these supports as temporary scaffolding to be removed once the delegation completes rather than permanent infrastructure.
When building behavioral chains through habit stacking, automate each link fully before adding the next link to the chain, because stacking unautomatic behaviors creates a pile of effortful tasks that collapses under executive function depletion rather than a self-sustaining sequence.
When someone shares expertise or makes a recommendation, separate your evaluation of their reasoning from your evaluation of their credentials by asking 'Would I find this compelling if it came from a low-status source?'
Before acting on AI-generated conclusions, apply the defense test: 'Could I defend this conclusion without the AI's output? Do I understand the reasoning well enough to identify where it might be wrong?'—if not, do the cognitive work before proceeding.
Before adopting anyone else's recommendation, apply the accountability check: 'Am I willing to own this decision as though it were entirely my own?'—if you would deflect blame to the source upon failure, you have not processed the input as influence.
When you notice an objection dissolving before you voice it—not because someone addressed it, but because the social cost seems too high—recognize this as the compliance instinct activating and deliberately externalize the concern in writing before deciding whether to voice it.
During emotionally charged disagreements in close relationships, write down your actual position before the conversation and return to it afterward to check whether changes were driven by persuasion or anxiety relief.
During dissent in group settings, when you are the sole opposing voice, explicitly state both your position and one condition under which you would change your mind to signal intellectual humility alongside independence.
In close relationships, frame disagreements using 'I think/want/believe' language rather than 'don't you think' or 'most people' formulations to take explicit ownership of your position.
Before a difficult conversation in a close relationship, externalize your position in writing with three components: what you think, what you're willing to change, and what you're not willing to change.
When you possess domain-specific expertise relevant to a decision, voice your dissenting assessment even when it conflicts with hierarchical authority, because withholding situated knowledge creates epistemic fragility and constitutes professional negligence.
Frame professional dissent as questions rather than assertions when building credibility, because questions activate information-sharing while reducing social cost and allowing decision-makers to reach conclusions themselves.
Build authority to dissent through demonstrated competence and calibrated prediction tracking, because self-authority without proven expertise is indistinguishable from arrogance and carries no credible weight.
Eliminate variable ratio reinforcement triggers by turning off all push notifications from social platforms, because every notification is a prompt delivered at a moment of susceptibility designed to initiate engagement without triggers, breaking the reinforcement schedule.
Audit your beliefs for algorithmic origins by identifying which positions trace primarily to repeated exposure within algorithmically curated environments, because beliefs acquired through engineered exposure rather than deliberate inquiry have not been subjected to sovereign epistemic standards.
When using AI systems, maintain explicit separation between input generation (AI-assisted) and judgment synthesis (human-retained), because outsourcing the integration function constitutes epistemic abdication regardless of input quality.
Conduct quarterly authority audits for one domain at a time using a structured format that records source, domain, basis of trust, scope of expertise, last verification date, and delegation level.
Trace each trusted source to whether trust stems from demonstrated expertise, emotional resonance, social proof, first-encounter effects, or algorithmic repetition to detect unexamined authority delegations.
Before making significant decisions, construct a diverse input set that includes at least one person likely to disagree with your current leaning, one with direct domain experience, and one outside the domain who might see what insiders miss.
Seek people who genuinely disagree with your position rather than assigning devil's advocate roles, because only authentic dissent triggers the broader cognitive processing that improves judgment quality.