I have been exploring creative writing with an AI partner for the past year. Recently my efforts have been focused on first person writing. I use AI as a writer, a lab, and a lab partner. I am a retired engineer so I favor ideas that are testable over ones that aren't. AI has has allowed me to test abstract concepts. My exploration of first person writing has really been an interesting journey.
As an example, I recently wrote a twenty-two chapter book about a man and two cats living in an apartment. I created an AI writing voice by having AI sample the works of several good writers and created a composite voice. The resulting voice held across 320 kindle pages. The work went fast with very little drift and when tested against my own judgement and AI critique tools was judged good.
I took the win and moved on. the problem occurred on my next effort. I had difficulty maintaining the same quality and speed. Drift became a real problem. I began searching for a solution and realized that the reason the original voice held was because of the initial conditions set on the writing. The man was the first person narrator. The two cats were the stimulus, and the apartment, the environment. The initial conditions were very constrained. AI was forced to operate in a tight space with very little room to drift.
To prove this I created a Jack Kerouac voice map and had AI write in Jack's voice under the same initial conditions as I used for the cat book. The surprise was that the writing was almost identical in tone and voice. The Kerouac persona was gone. After some other testing with changing stimuli and environment I concluded that a voice is determined by both, and moves in writing space by the conditions set by the input anecdote.
I could test this using AI as a lab and partner. I took the voice and changed the initial conditions. Kerouac writing about a man camping in the wilds or in the cockpit with Chuck Yeager in his successful attempt at breaking the sound barrier. Or even writing about a cat sleeping.
The AI text changed and tone and voice changed accordingly. I now had a lab to test the abstract idea of first person writing.
From there I made the leap that first person writing is actually modeling the thoughts of the narrator. I came up with three states, the reflective, inner facing, and the reactive outer facing and the accounting layer. The ultimate criteria is survivability. The reflective state is not bound by time or space and the reactive lives in the instant and stimulus from the environment. The accounting layer judges the results and feeds back to the reflective.
By trial and error I came up with transition rules and tested their validity. Easy, I just read the AI writing and decided if the output fit the conditions and stimuli specified by the "Write me a story about a man walking down a busy street and meets an old friend". Eventually I had the transitions and a writing specification to feed AI before working with him to create a story.
Now even more interesting. After locking down the writing rules, I made another leap. What I had worked for first person writing but what if I extended the work for the next effort and applied it to real people or groups of people, or anything else depending on human input. I started with the idea of the tribes leaving the old by the road to die. Now current society judges this as cruelty. I had a different idea. The elders in ancient tribes were treated with the utmost respect and the keepers of wisdom. What if they made that choice and it was a sacrifice of the highest order. They die to save the children, their grandchildren. I would judge this as a reactive act that ensured the survivability of the group. And as an aside authors who write about this treat the scene as sad but seldom do the elders rant and rave about their fate.
Nothing radical or new, except that I could test my ideas, that was new.
So in short, I modeled the human as reflective, reactive, and accounting layers all working together to ensure survivability. The reflective learns from the outside world and sets the constraints, rules and boundaries on the reactive, the account layer judges the reactive and feeds its judgement back to the reflective and the system adjusts.
I add ideas as they occur, mostly in the accounting layer, starvation of the children, failure to pivot, usage of excess capacity, The model doesn't hedge, everything is judged as 1/0, if something is a threat and the action is reflective rather than reactive, its a failure. I then expanded the method to groups, societies, countries, even presidential administrations. It seems that I was on the write track. It's an interesting journey.
AI is magic. This is just a bit about how I am using it. It has helped me to understand others, and understand myself.
Follows are short examples of first person AI writing with my rules and an output of the AI survivability analysis method.
First a test is a gunfight in the old west, AI written and analyzed:
The dust is a fine, gold powder suspended in the stagnant air of noon, coating the brim of my hat and the dry skin of my knuckles as I watch the man thirty paces away. He moves with a slow, deliberate hitch in his gait, his shadow a black puddle beneath his boots, and I find myself wondering if he remembers the card game in San Antonio or if I am just another silhouette blocking his path to the next bottle of rye. There is a strange, detached clarity in the way the sunlight glints off his silver collar stud, a tiny star against the sweat-stained linen of his shirt that seems more real than the heavy weight of the Colt hanging at my hip. I can feel the heat radiating off the adobe walls, a dry, baking pressure that makes the silence of the town feel like a physical thing, a glass dome waiting for the first sharp sound to shatter it into a thousand jagged pieces.
His thumb hooks into his gun belt.
My knees lock, my center of gravity shifting downward as my peripheral vision swallows the empty storefronts and focuses entirely on the slight tension in his forearm.
The world shrinks to the width of a leather holster.
He’s younger than I thought, his eyes narrowed against the glare, a kid looking for a legend to bury so he can take the name for himself.
A horse whinnies in the livery stable, the sound sharp as a whip-crack.
I don’t wait for him to find his rhythm; my palm meets the checkered grip of the revolver, the metal warm from the sun, and I draw in one fluid, mechanical motion that bypasses thought entirely.
The barrel clears leather just as he begins his sweep.
The first shot is a heavy, concussive punch that kicks back into my wrist, the roar of the black powder drowning out the wind and the ghost of the San Antonio memory.
He stumbles, his own shot plowing a furrow into the dirt three feet to my left.
I don't breathe until the smoke clears enough to see him slumped over his own shadow.
The street is quiet again.
Execution Analysis
Mode Transition Boundary: When his thumb hooked into his belt, the paragraph terminated immediately to signal the shift from Reflective to Reactive mode.
Cognitive Visibility: The long opening paragraph represents the "expanded" time of the standoff, while the gunfight is rendered in fragmented, single-sentence paragraphs to reflect Survival Priority.
Interruption Without Completion: The reflection on the San Antonio card game was cut off mid-thought by the environmental demand of his movement.
Here is a pressure test of an early version of the survivability method, its way more advanced now:
Using the three-layer rules (reactive / reflective / accounting → survivability), you can “pressure test” failed societies by asking one question per layer: what did they do under load, what did they learn (if anything), and what forced contact with reality (or got bypassed).
Rome, late Western Empire (3rd–5th century). Reactive: repeated emergency responses—short-term military fixes, debasement, ad hoc taxation, power shifting to whoever could hold territory. Reflective: reforms existed but rarely expanded capacity; the system kept patching the same boundary failures (succession instability, fiscal base erosion, frontier overload). Accounting: reality arrived as lost revenue and lost control of logistics; when the tax and recruitment base can’t fund the reactive layer, the state becomes a shell. Survivability fails when reaction is always “crisis mode” and reflection can’t produce structural constraints that reduce crisis frequency.
Weimar Germany (1919–1933). Reactive: governing by coalition fragility, street violence, and repeated emergency measures; the system’s reaction increasingly became “manage instability” rather than “restore legitimacy.” Reflective: the political class did not compress lessons into enforceable constraints that reduced polarization and delegitimization. Accounting: hyperinflation and later mass unemployment were hard accounting events; once the public’s lived consequence contradicts institutional narratives, the reflective layer loses authority and reaction gets captured by actors promising decisive action. Survivability fails when accounting shock is large and the reflective layer cannot translate it into trusted reforms.
Soviet Union (late 1970s–1991). Reactive: maintain the machine—security, subsidies, and control—while productivity and innovation lag; reaction stayed “stable” but increasingly irrelevant to survival requirements (economic competitiveness, legitimacy). Reflective: reforms (perestroika/glasnost) were not cleanly compressed into a new executable operating system; they destabilized the old constraints without installing durable new ones. Accounting: budget strain, shortages, and declining legitimacy were the real ledger; once republic-level accounting diverged from center-level story, the union could not enforce a single truth-loop. Survivability fails when reflection loosens constraints faster than it builds replacement constraints, and accounting fractures into competing ledgers.
Yugoslavia (late 1980s–1990s). Reactive: under economic and political load, reaction devolved to factional security and ethnic mobilization; survival function shrank from “shared state” to “group dominance.” Reflective: no credible mechanism could compress shared learning across groups; reflection became propaganda. Accounting: violence becomes the accounting system—force replaces institutional consequence. Survivability fails when the shared accounting layer collapses; without a common ledger, there is no shared truth and no stable reflective feedback.
Venezuela (2010s–). Reactive: preserve regime stability via controls, patronage, and coercion while productive capacity erodes; reaction continues, but it is tuned to internal survival of the ruling system rather than societal survival. Reflective: policy learning is suppressed; failures are narrated as external sabotage instead of being compressed into constraints that restore capacity. Accounting: inflation, shortages, emigration are the ledger; if accounting is politically survivable, drift can persist a long time, but societal survivability degrades. This is the “slow death” mode: the state can remain, while society fails.
Somalia (state collapse, 1991–). Reactive: central reaction failed; security and service provision localized to clans/regions. Reflective: learning happens, but at smaller scales; the “society” becomes multiple societies. Accounting: consequence is immediate (security, trade routes, famine risk), so local systems that align reaction with accounting can survive even as the national entity fails. Survivability fails at the top when there’s no mechanism to recompress local truths into a shared national operating system.
One meta-result falls out across all of these: collapse is usually not “bad reaction.” It’s “reaction divorced from accounting” and “reflection that does not expand reaction.” When the shared accounting layer fractures (multiple ledgers, multiple truths), “society” stops being one sealed world. Your “2+ worlds” principle becomes literal: competing sealed worlds form inside the same borders, and the original society fails by definition.
United States (roughly 1789–1860).
Maturing window: the early republic through the mid-19th century, when federal institutions, taxation/finance, and conflict-handling capacity increased without constant regime rupture. Reactive capacity expanded (administration, commerce, territorial governance). Reflection compressed into executable constraints (Constitutional mechanisms, courts, budgeting, professionalization). Accounting was harsh and real (credit crises, elections, wars) and generally fed back into institutional adjustment rather than being denied. Grade: PASS for maturation in that window. Failures existed (slavery is an existential contradiction), but the system’s capacity growth and correction loops were real until the contradiction overwhelmed it.
West Germany (Federal Republic), roughly 1949–early 1970s.
Maturing window: post-1949 reconstruction through the “economic miracle” period and consolidation of democratic institutions. Reactive: rapidly improving state capacity and economic execution. Reflective: lessons from collapse were compressed into durable constraints (constitutional design, checks, party structure, integration with European institutions). Accounting: hard external constraints (markets, allies, reconstruction limits) prevented narrative-only solutions; performance was measurable and punished quickly. Grade: PASS.
Japan (Meiji period), roughly 1868–1912.
Maturing window: state modernization and institutional build-out. Reactive: quickly increased administrative and industrial capacity. Reflective: imported learning was compressed into rules, education, bureaucracy, and military/industrial systems. Accounting: competitive pressure from external powers forced consequence; capacity had to become real. Grade: PASS for maturation in that window (separate from later imperial trajectory).
South Korea, roughly early 1960s–1997.
Maturing window: rapid industrialization through democratization (late 1980s) up to the Asian Financial Crisis. Reactive: expanding execution capacity (industry, export infrastructure). Reflective: policy learning did compress into institutions over time (eventually including democratization), though often late and under stress. Accounting: global markets imposed hard consequences; 1997 was a severe accounting event that forced restructuring rather than total narrative denial. Grade: PASS (with a warning: early period often relied on coercive reaction; maturation was real but uneven until governance caught up).
A quick way to see “maturing” vs “failing”,maturation is when crises produce constraints that permanently raise the reactive boundary. Failure is when crises produce stories, scapegoats, or churn that leave the boundary unchanged.
So this article illustrates the amazing ideas a man and ai can achieve when working as equal partners. And using AI as lab and lab partner permits testing the abstract. That's an advancement far beyond prompts!
