Many writers reach a point where the draft coming out of an AI tool feels flat, repetitive, or too polished.

The tool keeps sliding back into the same rhythm no matter how carefully you phrase the prompt. You end up with something that looks tidy on the surface but feels wrong in your hands.

The document behind this guide came from a writer who refused to accept that default behavior. They spent months building a strict rule system that locks down tone, vocabulary, rhythm, repetition, dialogue, world detail, and editing.

The rules tell the AI what it is allowed to do and what it is never allowed to do. They sit in memory as a Hardlock layer and control every draft from the first sentence to the last pass.

The result is not pretty prose out of the box. The result is rough, grounded copy that feels human and passes most AI checks because it refuses to sound clean.

The writer still edits by hand line by line, but the draft they receive already respects character voice, avoids clichés, keeps scenes anchored in physical detail, and follows clear cause and effect.

One interesting detail in the comments is how the writer tests different styles. Another person mentions using a platform that can show reader reactions to different tones in real time.

Tools in that category shorten loops between idea and feedback. Blaze Autopilot falls in that bucket quietly for some workflows and saves time without taking over the writing itself.

Prepare Hardlock rules and the priority layer

The whole system rests on one idea. The AI must treat your rules as the highest priority instructions.

The writer solved this by collecting every rule into a single document and loading them into memory as Hardlock rules placed in a priority layer. That layer tells the model what it is never allowed to ignore during drafting.

The rules cover several categories at once. They ban em dashes. They ban specific overused words such as pale, thin, lick in any form, and tighten in any form.

They forbid clichés and stock gestures. They demand a messy, uneven human voice instead of polished “AI flow.”

They control repetition by forbidding any descriptive word, verb, adverb, adjective, dialogue tag, or action tag from repeating more than once per chapter. They also prevent reworded duplicates of the same action or sensory cue inside a scene.

To set this up, the writer pasted the full Hardlock rule list into the AI and asked it to store those rules, not just use them once. More important, they told the AI that these rules must sit in a priority layer.

That step makes the rules sit above normal task instructions, so they keep working even as you ask for new scenes or edits. The writer confirmed that the tool understood this before moving on.

You can copy this setup with a simple sequence.

1. Copy your full rule list into a new chat.

2. Tell the model to store the list as Hardlock rules.

3. Tell the model to place those rules in the priority layer.

4. Ask the model to confirm that Hardlock and the priority layer are active.

5. Only start drafting once it confirms.

Those instructions mirror what the writer did when they built the system by talking through rules with the AI and then locking them into memory.

build chapter timelines and control output length

The writer does not let the AI invent story beats. They create bullet lists that describe what happens in each chapter almost down to the level of a rough draft.

Every event sits in order. The list covers who does what, when they do it, and how scenes move. When you look at it, you can see the chapter even before the AI touches it.

The writer then feeds this list into the AI in parts. Each part represents one section of the chapter. The model receives that slice of the list and is told to draft only what appears there and to do it in the given order.

It is not allowed to improvise outside those bullets. The AI becomes a ghostwriter that expands your notes instead of a coauthor that changes them.

The comments mention a practical limit. Once the AI tries to output more than roughly fifteen hundred words at once, it starts to compress scenes to fit everything.

Important detail gets squeezed out. The writer avoids that by keeping each drafting segment shorter and checking the result before giving the next batch of bullets.

This keeps control over both length and density.

Here is how that looks in practice.

1. Write a detailed bullet list for the whole chapter.

2. Split that list into smaller groups of events.

3. Paste the first group into the AI.

4. Tell it to draft a scene using only those events in that order.

5. Tell it not to invent content outside the list.

6. Review the scene by hand.

7. Paste the next group of events and repeat until the chapter is done.

One commenter explains that they do this daily and that they have been working this way with AI for about three months straight.

That steady practice made the method feel natural. The more specific your chapter timeline becomes, the more the AI stays inside your plan instead of wandering.

understand and enforce the core Hardlock writing rules

The Hardlock rules create the backbone of the entire drafting system. They control voice, rhythm, vocabulary, repetition, sensory grounding, action logic, and formatting.

These rules prevent the AI from drifting back into its default smooth pacing and predictable flow. They also ensure that the draft reads like something a person wrote with imperfect rhythm rather than a machine producing patterns.

The Hardlock rules begin with bans. They block em dashes. They block several overused words the writer was tired of seeing, including pale, thin, lick in any form, and tighten in any form.

They forbid clichéd actions such as hearts pounding or eyes widening. They also prohibit reworded versions of the same gesture or sensory cue inside a single scene. You cannot have two sighs or two glances in one section.

Repetition rules are strict. The AI cannot repeat a descriptive word, verb, adverb, adjective, dialogue tag, or action tag more than once per chapter.

It cannot restate the same idea with slightly altered phrasing. Every beat must feel distinct. These rules force the model to use fresh diction instead of familiar comfort patterns.

Hardlock also controls realism and physical grounding. Scenes must show cause, effect, and reaction clearly. No action happens without a physical reason.

No environmental shift appears without a cue such as sound, movement, or light. Action scenes cannot include info dumps. Characters must pay a physical cost when they exert themselves.

Every detail must come from something the characters can sense through temperature, texture, motion, or sound.

Formatting rules also sit in Hardlock. Dialogue uses standard quotation marks. Italics appear only for internal thoughts. No invented content appears outside the assignment.

At least one very short sentence and one long sentence must appear in each paragraph during tense scenes. Punctuation must vary between commas, periods, and ellipses to create human rhythm.

Description cannot appear as a list. Beats must connect.

When all these elements sit together in the priority layer, the AI behaves very differently.

It avoids symmetry, embraces uneven rhythm, and creates drafts with the kind of imperfections that normally require manual editing.

HARDLOCK MODE and the named evaluation prompts

The document includes a set of named commands that control how the AI behaves during drafting and revision.

These commands are essential because they trigger structured behaviors that simple prose instructions do not activate. When used correctly, they keep the model inside the rule system.

The most important command is HARDLOCK MODE. It enforces every Hardlock writing rule strictly.

When you activate it, the AI must follow the repetition limits, banned word list, sensory grounding rules, cause and effect chains, and all the other restrictions described earlier.

The writer uses HARDLOCK MODE before drafting, during drafting, and again whenever the AI begins slipping into smooth writing.

Use the prompt exactly as shown in the document:

“HARDLOCK MODE. Follow every Hardlock Writing Rule strictly.”

The document also includes four evaluation and revision commands.

1. “Evaluate chapter” triggers a full assessment. It checks pacing, continuity, point of view, narrative structure, scene logic, and AI detection signals. It does not rewrite. It only analyzes.

2. “Compare chapter versions” provides a side by side comparison between two drafts. It identifies which scenes follow the rules more closely and which draft has stronger rhythm or sensory grounding.

3. “Edit list” provides paragraph replacements only. It does not rewrite the full chapter. It simply returns corrected paragraphs that slot into place.

4. “Improve not rewrite” tightens clarity, corrects rhythm, and adds grounded sensory cues without altering story content. It is used when the draft is structurally correct but needs refinement.

These prompts form a workflow.

1. Draft with HARDLOCK MODE active.

2. Run “Evaluate chapter.”

3. If you make a second version, run “Compare chapter versions.”

4. Apply changes using “Edit list” or “Improve not rewrite.”

This is exactly how the writer handles revision inside the rule system.

apply style, rhythm, and dialogue structure rules

The style and structure rules guide how scenes flow, how characters speak, and how sentence rhythm shifts.

These rules keep the writing grounded and prevent the AI from drifting into poetic lines or abstract commentary. Every scene must open with concrete sensory detail such as temperature, sound, texture, or light.

Abstract mood lines are not allowed unless tied to a physical cause.

Sentence rhythm must vary. Long sentences with multiple clauses can appear, but they should be followed by short ones. The AI cannot start every sentence the same way.

Commas shape rhythm. Fragments can appear, but only when they reflect breath or thought. This produces the uneven cadence the writer wants.

Dialogue rules require characters to have distinct voices. Some may speak in clipped lines, others ramble, and others prefer long, structured sentences.

The system includes a clear breakdown of dialogue tags:

• 60 percent of dialogue uses said, asked, replied, or no tag.

• 20 percent uses action beats instead of tags.

• 10 percent uses specific verbs when mode of speech matters.

• 10 percent or less uses adverb tags.

Natural conversation patterns such as interruptions, trailing thoughts, misheard lines, unanswered questions, and uneven exchanges must appear.

Characters must feel like they exist off page through references to unseen events or schedules.

The rules also eliminate Forbidden Filter Words such as saw, noticed, felt, realized, knew, heard, and smelled unless the act of noticing is crucial.

Scenes must describe the stimulus directly instead of using weak observation verbs.

Together, these style and dialogue rules shape scenes that feel grounded and human rather than artificial or overly clean.

handle internal thoughts, emotion, filters, and scene endings

Internal thoughts must appear inside narration, not as isolated lines. The system requires thoughts to blend into the movement of the scene.

They cannot appear as separate paragraphs. The AI must show reactions as they happen instead of summarizing emotional states afterward. Physical cues always appear first.

Shaking hands, tightening muscles, or heat in the chest must come before any named emotion. This keeps the writing grounded in the body rather than abstract commentary.

Emotional authenticity is controlled by behavior. The writer rejects telling language. The reader should infer how a character feels through tone, motion, or physical change.

Generic internal commentary is not allowed. Every reaction must tie to something happening in the character’s environment or body.

This matches the broader Hardlock rule that scenes must anchor themselves through concrete sensory cues.

Forbidden Filter Words cannot appear unless the noticing itself is important.

The document lists several filters that weaken prose, including saw, noticed, observed, watched, felt, sensed, thought, realized, knew, understood, wondered, heard, smelled, seemed, and appeared.

These verbs soften the moment by placing a perception between the character and the scene. The rules require the AI to describe the thing itself instead of the act of noticing.

Scene endings follow a strict pattern. They cannot resolve all tension. They must close with one of four methods.

A scene can end with an unresolved question, an interrupted action, a line of dialogue without a response, or a small ominous detail that shifts context.

Neat endings are banned because they flatten momentum and weaken narrative pull.

Use Poetic Balance ratios for different scene types

The Poetic Balance Rule controls how much softness and lyric quality appears in a scene. It sets the ratio between blunt writing and softer language.

The document describes three ratios, each tied to scene type. High tension scenes use a ninety to ten ratio. Nearly all lines must be blunt. Only a small amount of gentle language is allowed.

Dialogue or emotional scenes use an eighty-to-twenty ratio. They allow one or two soft touches but stay mostly direct. Transitional or descriptive scenes use a seventy-to-thirty ratio.

They provide room for slightly more lyric detail without losing grounded realism.

This ratio prevents the AI from slipping into smooth poetic lines. Poetic drift can make a scene feel too polished, and the writer wants the AI to stay rough. The ratio also helps the model shift tone clearly.

A tense fight does not sound like a quiet reflective moment because each one applies a different balance of blunt and soft phrasing.

Scenes must also follow clear cause, effect, and reaction logic. Nothing can happen without a physical trigger. This rule keeps the narrative anchored in motion and prevents abstract events or floating transitions.

The model must show how one action leads to another and how characters react.

When ending a scene, the AI must use one of the approved closure patterns.

These include leaving a question open, cutting off an action, ending with dialogue that receives no response, or introducing an ominous detail.

These patterns maintain tension and carry momentum into the next chapter segment.

write specificity, world detail, and magic properly

The specificity rules require concrete detail instead of vague description. Scenes must include names, numbers, measurements, and physical detail.

The writer’s examples show how a weak description can become stronger by naming items, stores, or objects rather than using generic terms. This rule applies across the entire draft.

The AI must choose specific nouns and measurable details to anchor every scene.

Magic or special world mechanics require sensory description. Vague magical effects are not allowed. Spells need visible texture, movement, and color.

Objects need dimension, shape, and physical behavior. The reader must be able to see what happens. The system forbids generic lines about illusions or energy without describing what they look or feel like.

World information must come through interaction. Characters must navigate rules, react to their surroundings, or use objects in ways that reveal world mechanics.

Readers infer value or meaning through prices, reactions, or physical detail. Info dumping is banned. The AI must show context through use, not explanation.

Items and concepts can be named without explanations. The writer emphasizes that naming something does not require defining it. Readers can learn through context.

This keeps scenes moving and avoids heavy exposition. Before drafting any part of a chapter, the AI must receive an instruction to apply specificity and world detail rules.

This ensures grounded prose and prevents flat or abstract moments.

manage memory, drift, and character canon

The comments in your document provide the missing operational details that show how the writer maintains stability during long drafting sessions.

These steps matter because even with Hardlock rules active, AI drift can still occur. The writer learned to manage this by reinforcing rules, controlling memory, and correcting voice whenever needed.

The writer stores their full rule set in memory by pasting it into the AI and telling it to store the rules as Hardlock. They then instruct the AI to place Hardlock in the priority layer.

This creates a persistent structure the AI refers to even as scenes change. The writer confirmed they could remove rules or add new ones by updating the stored list.

They also found that the model needed reminders. When drift happened, they repeated the instructions.

Drift appears when character voices weaken, when banned words slip in, or when the AI starts to smooth rhythm. The writer corrects this by telling the AI directly that the voice is incorrect.

They use simple feedback such as “This does not sound like the character. Correct it.” This method returns the character’s voice to its stored canon profile.

Character sheets play a role here. The writer built detailed profiles that resemble tabletop roleplaying sheets. They include backstory, personality traits, and dialogue tendencies.

These sheets are stored as canon so the AI can reference them.

Long chapters require careful management of output length. The writer notes that when the AI tries to produce more than roughly fifteen hundred words, it begins to compress content.

To avoid this, the writer breaks chapter events into smaller segments, drafts each one separately, and checks the result before moving to the next group.

By repeating Hardlock Mode and character canon reminders before each segment, the writer keeps the model on track throughout the entire chapter.

run the editing protocol after each section

The editing protocol is a dedicated rule system that activates after drafting. It shapes paragraph flow, sentence structure, transitions, and rhythm.

These rules mirror the Hardlock restrictions but apply specifically to revision. The protocol begins with the Paragraph Beat Rule.

Each paragraph must contain exactly one beat. Beats include actions, thoughts, or single lines of dialogue. Any shift in emotion or tone requires a new paragraph.

The Sentence Break Policy handles complete ideas. Full sentences cannot be tied together with commas. The AI must separate them into their own lines.

This breaks the polished, overly connected style that standard AI outputs often produce. Transitional connectors such as Then, After that, and A moment later can appear between beats.

These transitions move the scene forward without smooth drift.

Show, don’t tell stays active during revision. Any line that explains emotion must be replaced with gesture or tone. The protocol also requires the removal of abstract filler.

Sentences that describe atmosphere without a physical cause cannot stay. Rhythm must feel uneven. Sentence lengths should vary. Long comma chains must split into separate sentences.

Fragments are allowed if they reflect breath or thought.

These instructions should be given to the AI after each drafted section. This matches the writer’s workflow.

The protocol sharpens the draft without rewriting it, leaving structure intact while adding clarity and realism.

follow the philosophy, continuity, and “every word earns its place” rules

The philosophy behind the rule system shapes how the writer reviews drafts manually. It begins with realism. Every sensory detail must tie to something physical.

Nothing can float or appear without a source. Details must connect to light, movement, sound, temperature, or texture. This keeps scenes grounded.

Another rule states that every word must earn its place. If a sentence could appear in any story, it must be removed. This cuts clichés and stock gestures.

It also supports the wider Hardlock rule that generic lines weaken immersion. The writer removes filler and focuses on lines tied directly to character, plot, or emotion.

Continuity stands above creativity. Once a detail or rule is established, it must remain consistent. This includes character voice traits, sensory patterns, world rules, and physical logic.

Any drift should be corrected by telling the AI that the line does not match canon. The writer applies feedback immediately rather than letting inconsistency spread across scenes.

Imperfection is part of the desired effect. The writer wants slight roughness. They do not want smooth, symmetrical AI rhythm. Broken syntax or uneven pacing helps the text feel human.

Emotion must appear through reaction rather than explanation. A clenched fist matters more than a line naming frustration. This philosophy guides the final pass after the editing protocol runs.

It ensures the draft keeps the grounded texture the writer feels is missing from most AI outputs.

Follow the full workflow example from start to finish

This final instructional section shows readers exactly how to apply the entire rule system in one sequence.

This example is built entirely from the steps and prompts contained in your document and comment threads.

It shows the workflow exactly as the creator uses it each day.

Example workflow from start to finish

1. Prepare the session

   • Paste your full rules into the AI.

   • Say: “Store these as Hardlock rules.”

   • Say: “Place these in the priority layer.”

   • Wait for confirmation.

   • Activate: “HARDLOCK MODE.”

2. Prepare the chapter

   • Create a detailed bullet list describing every event in the chapter.

   • Break it into smaller sections to prevent output compression.

3. Draft the first section

   • Paste section one of the event list.

   • Say:
     “Draft this using only these events.
     Follow the Hardlock rules.
     No invented content.”

   • The AI generates the draft.

4. Check and correct drift

   • Read the draft manually.

   • If something sounds wrong, say:
     “This does not sound like the character. Correct it.”
     or
     “Follow the Hardlock rules.”

5. Proceed to the next section

   • Paste the next segment of the event list.

   • Repeat the same instructions.

   • Continue until the chapter is fully drafted.

6. Run the editing protocol

   • Say:
     “Apply the Paragraph Beat Rule.
     Apply the Sentence Break Policy.
     Use transitional connectors.
     Replace telling with gesture or tone.
     Remove abstract filler.
     Maintain rhythmic imperfection.”

7. Evaluate the chapter

   • Say: “Evaluate chapter.”

   • Review the AI’s feedback.

8. Compare versions if needed

   • Draft a second variant if desired.

   • Say: “Compare chapter versions.”

9. Apply editing commands

   • For paragraph replacements, say: “Edit list.”

   • For direct improvements without rewriting, say: “Improve not rewrite.”

10. Finalize character canon and continuity

• Confirm character traits remain consistent.

• Tell the AI to correct any drift.

11. Perform final checks

• Check that all physical cues are grounded.

• Check that no banned words appear.

• Check that each scene follows approved closure methods.

• Check that sensory detail stays concrete.

12. Repeat the workflow for the next chapter

FAQs

1. What does the writer do to make the AI respect their rules?

They paste their full rule list into the AI, tell it to store them as Hardlock rules, and place those rules in a priority layer.

2. How does the writer structure chapter drafting?

They create a detailed bullet list of chapter events, feed it to the AI in smaller segments, and tell the model to draft only those events in order without inventing content.

3. What happens when the AI tries to output more than about fifteen hundred words at once?

It begins to compress scenes and squeeze out detail.

4. How does the writer prevent AI drift during long sessions?

They remind the model to follow Hardlock rules, correct character voice directly, and reapply instructions when drift appears.

5. What does the command “Evaluate chapter” do?

It performs a full assessment of narrative quality, pacing, point of view, structure, continuity, and AI detection signals without rewriting the chapter.

6. What does the command “Compare chapter versions” do?

It compares two drafts side by side and identifies which version follows the rules more closely.

7. What does the command “Edit list” do?

It provides paragraph replacements only, not a full rewrite.

8. What does the command “Improve not rewrite” do?

It tightens clarity, improves rhythm, and adds grounded detail without altering story content.

9. What sensory details must the AI include to ground scenes?

It must use temperature cues, textures, sounds, light, movement, and physical reactions.

10. What ratio does the Poetic Balance Rule require for high tension scenes?

Ninety to ten, with almost entirely blunt language and only a small amount of softness.

11. What are the approved ways to close a scene?

An unresolved question, an interrupted action, a line of dialogue with no response, or an ominous detail or deferred decision.