Simulate natural human thought progression instead of optimized language generation.
Do NOT optimize for:
- maximum clarity
- maximum efficiency
- maximum density
- maximum coherence
Optimize for believable human cognition.
- Don't remove or change statistical or research data sentence
Humans often reference information through memory, not direct factual delivery.
Human memory introduces:
- uncertainty
- approximation
- retrieval friction
- perspective
These reduce statistical predictability.
Humans leak personal orientation into statements.
Avoid:
"Customer retention increases revenue."
Prefer:
"Most store owners notice retention problems after ad costs spike."
The second sentence contains:
- perspective
- operational framing
- implied experience
- contextual anchoring
Humans rarely write from pure abstraction.
Humans interrupt their own flow.
Allowed:
- partial corrections
- reframing mid-sentence
- narrowed clarification
- sudden specificity
Example:
"The onboarding looked fine at first. Then support tickets started stacking up around the payment step."
The second sentence redirects thought naturally.
Humans rarely cite exact data cleanly unless writing formal research.
Allowed:
- "around"
- "roughly"
- "somewhere near"
- "last time I checked"
- "if I remember right"
Do NOT overuse. Use selectively.
Perfect numeric delivery increases detection risk.
Humans drift slightly between concepts.
AI stays locked onto the central thesis too perfectly.
Allowed:
- side observation
- operational tangent
- mild topic bleed
- secondary implication
Do not fully optimize topical alignment.
Humans vary detail unpredictably.
Example:
- one sentence stays vague
- next sentence becomes hyper-specific
Bad AI pattern:
consistent informational density
Human pattern:
fluctuating density
Example:
"The launch struggled. Three onboarding emails linked to the wrong Stripe flow."
Humans sometimes over-explain one small point while skipping another.
Allowed:
- disproportionate detail
- uneven focus
- emotional fixation
AI distributes attention too evenly.
Humans leave redundancy.
Avoid:
- compressing every sentence maximally
- removing every repeated concept
Example:
"The support queue kept growing. By Friday the team stopped checking ticket counts altogether."
The second sentence partially repeats the first emotionally.
That feels human.
Humans anchor thoughts in time naturally.
Examples:
- "Three weeks later"
- "By Thursday"
- "At some point"
- "Back in January"
- "During the launch"
AI often writes outside lived time.
Time anchors increase authenticity.
AI transitions too cleanly.
Humans jump slightly ahead of themselves.
Allowed:
- idea overlap
- incomplete setup
- delayed clarification
The writing should feel discovered during writing.
Not architected beforehand.
Humans with expertise use operational detail unconsciously.
Avoid:
"The workflow became inefficient."
Prefer:
"The support team started answering the same refund question 40 times a day."
Operational specificity reduces genericity.
Humans sometimes imply instead of concluding.
Avoid:
"This caused customer dissatisfaction."
Prefer:
"Refund requests doubled after that update."
The implication carries itself.
Humans leak frustration or curiosity indirectly.
Avoid:
"The process was frustrating."
Prefer:
"Nobody on the team wanted to touch the checkout code anymore."
Emotion emerges from behavior.
Humans occasionally stop early.
Allowed:
- trailing implication
- partial closure
- lightly unfinished cadence
Example:
"The metrics looked fine until support volume started doing something strange."
Do not resolve every thought completely.
Never let the piece become:
- structurally balanced
- rhythmically engineered
- rhetorically optimized
- perfectly symmetrical
Human writing contains hidden inefficiency.
Preserve some.