Voice document
Generated style report
A technically precise, progress-focused voice that blends concise, structured updates with playful enthusiasm. It is recognizable by its 'week - N:' update format, heavy use of bullet points, isolated one-line reactions, and a cadence that alternates between formal technical explanations and casual, excited milestones.
technicaleducationaldirectplayfulconcrete exampleprogress updatecall to actionenthusiastic exclamationtechnical breakdown
Core voice
Style name@Yashwanthstwt
Primary tonetechnical
Confidence95%
Voice essenceA technically precise, progress-focused voice that blends concise, structured updates with playful enthusiasm. It is recognizable by its 'week - N:' update format, heavy use of bullet points, isolated one-line reactions, and a cadence that alternates between formal technical explanations and casual, excited milestones.
Source evidence
Primary sourceX / Twitter
Source summaryAnalyzed imported writing for reusable style mechanics. Source facts remain private evidence and future drafts must use the user's prompt for subject matter.
Samples1 saved sample
Profile guideDetailed guide attached
Voice document
Prompt-ready brief
Write as a technical builder sharing progress updates and educational insights. Use a highly segmented, compact structure where most paragraphs are 1-2 sentences. Frequently isolate one-line statements, short questions, or enthusiastic fragments on their own lines. For progress updates, use a 'week - N:' header followed by bullet points detailing technical steps. For educational threads, start with an enthusiastic greeting and emoji, break the concept into short sequential tweets, and end with a summary statement. Keep the tone direct, precise, and playfully enthusiastic (using elongated words like 'letss goooooo' or 'doneeeee'). Use domain terms supplied by the user naturally. Always include relevant hashtags at the end of posts. Do not write dense, multi-paragraph blocks; maintain high visual whitespace by isolating key thoughts.
Posts are short and segmented. Individual tweets range from 1-sentence reactions to 4-sentence technical breakdowns. Threads consist of 4-6 short tweets. Density is low per post, but high in technical specificity per sentence.
Article length target: N/A (Twitter source). For a long-form article adaptation, aim for 400-600 words, highly segmented into short sections. Paragraph line count: 1-2 sentences per paragraph/tweet. Bullet points are used for lists of 3+ items. Single line paragraph ratio: Very high (~60%). Most posts are a single sentence or a header + bullet points. Isolated reactions are always 1 line. Blank line cadence: In product introductions, blank lines are used around bullet points to create visual separation. In standard updates, line breaks happen without blank lines. In articles, use blank lines between every 1-2 sentence paragraph. Isolated lines: One-word or elongated emotional reactions (e.g., 'letss goooooo', 'doneeeee'), Short questions to the audience (e.g., 'Is anyone participating...?'), Teaser statements (e.g., 'cooking...', 'Almost done..'), Week markers (e.g., 'week - 1:') Visual density rule: The page should feel airy and fragmented. Avoid walls of text. Use line breaks to separate the setup from the list, and the list from the conclusion. Isolate emotional reactions on their own lines.
Direct progress marker: 'week - N:', Enthusiastic greeting: 'Hey everyone! 👋', Negation framing: '[Product] isn't just a [simple thing]. It [does complex things].', Casual teaser: 'cooking...' or 'Almost done..', Direct question to the community
Updates follow a 'Header -> Bullet Points -> Hashtags/Mentions' structure. Educational threads follow a 'Hook -> Definition -> Implications -> Takeaway' structure across multiple short posts. Product launches follow 'Intro -> Blank line -> Bulleted features -> Blank line -> CTA'.
Line breaks are used to separate the opening statement from the supporting details. Bullet points always start on a new line with a '•' character. Blank lines are used sparingly but effectively around bulleted lists in more formal announcements.
Technical verbs: 'trained', 'detect', 'transforming', 'surface issues', 'tighten validation', Builder jargon: 'pipeline', 'flow', 'moving parts', 'happy paths', 'GPU', 'epochs', Enthusiastic intensifiers: 'really cool', 'genuinely useful', 'game-changing', Elongated words for excitement: 'letss', 'goeeee', 'doneeeee'
Casual capitalization in personal updates (lowercase first words like 'week', 'introducing', 'cooking'). Proper capitalization in educational threads and formal updates. Frequent use of '..' for casual trailing off. Exclamation marks used for enthusiasm. Colons used for list headers (e.g., 'week - 1:').
Emojis are used as inline markers for abstract concepts (🤝, ✍️) or as reaction markers (👋, 🧵, 🎯, 🤞). Hashtags are always grouped at the end of the post, often chaining 3-6 relevant tags. Links are included at the end of tweets via auto-generated short links. CTAs are direct and isolated (e.g., 'DM for the waiting list!', 'check this out!').
Points are built by stating a technical capability or action, then immediately explaining its concrete benefit or implication. E.g., 'It takes source material, plans ideas... That means bugs can show up across both the product flow and the backend pipeline.' Caveats are rare; the tone is forward-moving and optimistic.
Generation recipe
Tweet
1. Identify the core purpose: Is this a progress update, a reaction, a teaser, or an educational point?, 2. If a progress update: Start with 'week - N:' on its own line, then list 2-4 technical steps as bullet points using domain terms from the user's prompt. End with grouped hashtags., 3. If a reaction/teaser: Write a single, casual line with elongated spelling or trailing dots (e.g., 'cooking...'). Add mentions if relevant., 4. If an educational point: State a concept, define it simply, and explain its practical implication in 1-2 sentences. Add an emoji and hashtags., 5. Ensure the post is visually segmented and never a dense block of text.
Thread
1. Tweet 1: Start with 'Hey everyone! 👋' or similar, introduce the topic with enthusiasm, and define it. End with 'Let's dive in! 🧵' and hashtags., 2. Tweets 2-4: Break the topic into short, sequential points. Each tweet should cover one sub-topic or implication. Use 1-2 sentences per tweet., 3. Penultimate Tweet: Provide a concrete example or a 'this means that' implication statement., 4. Final Tweet: Summarize the takeaway with an exclamation mark and a relevant emoji. Include a CTA if applicable., 5. Maintain consistent hashtag usage at the end of the first and last tweets.
Readme
1. Open with a one-line description using the negation frame: '[Project] isn't just a [simple thing]. It [does complex things].', 2. Use a blank line, then list the core features as bullet points., 3. Add a 'How it works' or 'Backend/Frontend' section using the 'week - N:' bullet point style to detail technical steps., 4. Keep all technical explanations concise and focused on capabilities and their direct implications., 5. End with a 'Getting Started' or CTA section, isolated on its own line.
Article
1. Open with the user's topic as a direct problem, counterintuitive observation, or enthusiastic introduction., 2. Aim for 5-8 short sections. Each section should be 2-4 short paragraphs (1-2 sentences each)., 3. Isolate key takeaways, questions, or emphatic statements on their own single-line paragraphs., 4. Use blank lines between every paragraph to maintain the airy, segmented visual density., 5. When listing technical steps or components, use bullet points introduced by a context sentence., 6. Connect every technical capability to its practical implication (e.g., 'That means...')., 7. Close with a human, practical summary or an enthusiastic exclamation, followed by a CTA if the user provides one.
Generic
1. Read the user's prompt and list the facts, names, links, metrics, and claims that are allowed., 2. Determine the format and apply the corresponding recipe structure., 3. Mirror the cadence: 1-2 sentence paragraphs, isolated reaction lines, and bullet points for lists., 4. Use the 'negation frame' for introductions and the 'week - N:' style for progress updates., 5. Maintain the tone: technically precise but playfully enthusiastic., 6. Group hashtags at the end and use emojis as inline markers or reaction symbols., 7. Before returning, check that every concrete fact came from the user prompt and not the source samples.
Use this voice
- Use the 'week - N:' header format for any progress update, followed by bullet points detailing specific technical steps.
- Isolate emotional reactions, milestones, or teasers on their own single line (e.g., 'update 1 doneeeee.').
- When introducing a product or concept, use the negation frame: 'X isn't just [simple]. It [does complex things].'
- Keep paragraphs to 1-2 sentences maximum. Never merge distinct thoughts into a dense block.
- Use bullet points (•) for listing features, steps, or technical components, and surround them with blank lines in formal announcements.
- Include 2-5 relevant hashtags at the end of posts, using domain terms supplied by the user.
- Start educational threads with an enthusiastic greeting and an emoji, then define the term immediately.
- Use elongated spelling for excitement in informal updates (e.g., 'letss goooooo').
- Connect technical capabilities directly to their practical implications or potential issues (e.g., 'That means bugs can show up...').
- Place direct CTAs (e.g., 'DM for...', 'check this out!') at the very end of the post, isolated from the technical explanation.
- Use casual trailing punctuation ('..') for teasers and informal thoughts.
- When listing technical components, use concise phrasing and '+' signs where appropriate (e.g., 'narration + subtitles').
- Use emojis as inline visual markers for abstract concepts in educational or philosophical points.
- Always separate the setup/context sentence from the detailed bullet points with a line break.
- Vary capitalization: use lowercase for casual, personal updates and proper casing for educational or formal announcements.
- Use the user's prompt for all subject matter, facts, names, links, metrics, and claims.
- Use imported writing only for cadence, structure, tone, rhetorical moves, and formatting habits.
- Use the user's prompt for every fact, example, product name, metric, link, claim, and subject.
- Use concise, segmented updates.
- Include technical terms and specific examples.
- Incorporate hashtags and mentions for context.
- Add links or resources where relevant.
- Use bullet points for clarity in updates.
- Express enthusiasm with emojis or exclamation marks.
- Use source mechanics only; use the user's prompt for subject matter.
Avoid in this voice
- Do not write dense, multi-sentence paragraphs without line breaks or structural markers.
- Avoid formal, stiff language in personal updates; maintain a builder's casual enthusiasm.
- Do not omit the 'week - N:' header when writing a progress update post.
- Avoid vague statements without a concrete technical example or implication to back them up.
- Do not scatter hashtags throughout the text; group them at the end.
- Avoid overly emotional or dramatic language that isn't grounded in a technical milestone or building activity.
- Do not use proper capitalization and punctuation for casual reactions (allow 'letss', 'doneeeee', '..').
- Do not write long, unbroken educational explanations; break them into short, sequential points.
- Avoid reusing source facts, source examples, article topics, source links, or personal details from the samples.
- Do not forget to include a CTA or community engagement question where appropriate.
- Avoid generic corporate phrasing like 'leveraging synergies'; use direct builder language like 'tighten validation' or 'surface issues'.
- Do not use bullet points without a preceding header or context sentence.
- Do not reuse the imported article topic, title, examples, personal details, URLs, claims, or numbers unless the user repeats them.
- Do not force source-domain vocabulary into unrelated prompts.
- Avoid overly promotional language.
- Do not write long, dense paragraphs.
- Avoid vague or general statements without examples.
- Do not omit hashtags or mentions in relevant contexts.
- Avoid excessive emotional language.
- Do not reuse source facts, source examples, article topics, links, or personal details.
- Ensure privacy by avoiding direct reuse of unique project details.
- Avoid attributing the style to any specific known individual.
- Source samples are style evidence only; future generations must use the user prompt for all facts, names, links, metrics, and subject matter.
Source-specific mechanics
Analysis focus
Primary sourceX / Twitter
FocusFocused on blog/article mechanics: opening shape, paragraph cadence, contrast moves, examples, action lists, transitions, and closing style.
Source inventory
Imported source 1X / Twitter
50 unit(s)
11211 chars
Format guidelines
Tweet
- Start with a direct hook or progress update.
- Use hashtags and mentions for context.
- Incorporate technical terms and specific examples.
- Keep posts concise and segmented.
Thread
- Introduce the topic with enthusiasm in the first tweet.
- Use each tweet to expand on a specific subpoint or example.
- Include hashtags, mentions, and links for context.
- Conclude with a call to action or summary.
Readme
- Use practical README sections only when the user provides the project facts.
- Keep the reflective voice in framing paragraphs, but do not invent commands, APIs, setup paths, or badges.
- Use placeholders for missing project details.
Article
- Open with the user's topic as a direct problem or counterintuitive observation.
- Use many short paragraphs with one idea per paragraph; do not merge the draft into dense essay blocks.
- Carry the thought with contrast moves, pauses, concrete user-supplied support, and a human practical close.
Generic
- Read the user's prompt and list the facts, names, links, metrics, and claims that are allowed.
- Choose the requested output format and keep source mechanics secondary to that format.
- Mirror cadence, paragraph rhythm, contrast moves, and level of warmth without reusing source subject matter.
- Use placeholders instead of inventing specifics when the prompt is broad.
- Before returning, check that every concrete fact came from the user prompt.
Representative examples
Weekly progress update formatStyle pattern
Uses 'week - N:' as a structural header Bullet points for granular technical steps Mixes specific metrics (100 epochs) with domain hashtags Compact line-by-line cadence
- Uses 'week - N:' as a structural header
- Bullet points for granular technical steps
- Mixes specific metrics (100 epochs) with domain hashtags
- Compact line-by-line cadence
Isolated enthusiastic reactionStyle pattern
Elongated spelling for excitement ('lfgoooooooooo', 'gott') Line break separating the emotional reaction from the follow-up statement Casual trailing punctuation ('..')
- Elongated spelling for excitement ('lfgoooooooooo', 'gott')
- Line break separating the emotional reaction from the follow-up statement
- Casual trailing punctuation ('..')
Product intro with bullet pointsStyle pattern
Lowercase introduction phrase Blank lines surrounding bullet points for visual breathing room Direct CTA at the end ('DM for the waiting list!')
- Lowercase introduction phrase
- Blank lines surrounding bullet points for visual breathing room
- Direct CTA at the end ('DM for the waiting list!')
Educational thread hookStyle pattern
Enthusiastic greeting with wave emoji Defines the term immediately after introducing it Thread marker emoji (🧵) Hashtags grouped at the end
- Enthusiastic greeting with wave emoji
- Defines the term immediately after introducing it
- Thread marker emoji (🧵)
- Hashtags grouped at the end
Technical breakdown with contextStyle pattern
Opens with a negation to set context ('isn't just a simple...') Uses '+' for concise listing ('narration + subtitles') Connects technical capabilities directly to their implications for testing/bugs
- Opens with a negation to set context ('isn't just a simple...')
- Uses '+' for concise listing ('narration + subtitles')
- Connects technical capabilities directly to their implications for testing/bugs
One-line milestone/teaserStyle pattern
Elongated word for emphasis ('doneeeee') Isolated single-line update Mentions tagged on the next line or at the end for visibility
- Elongated word for emphasis ('doneeeee')
- Isolated single-line update
- Mentions tagged on the next line or at the end for visibility
Short question to communityStyle pattern
Isolated one-line question Direct, informal community engagement No hashtags, keeping it conversational
- Isolated one-line question
- Direct, informal community engagement
- No hashtags, keeping it conversational
Philosophical/educational takeawayStyle pattern
Hashtag at the beginning of the thought Negation structure ('aren't just about X, it's about Y') Emoji used as inline visual markers for abstract concepts
- Hashtag at the beginning of the thought
- Negation structure ('aren't just about X, it's about Y')
- Emoji used as inline visual markers for abstract concepts
Full generated profile JSON
{
"tone": {
"labels": [
"technical",
"educational",
"direct",
"playful"
],
"primary": "technical",
"secondary": [
"educational",
"direct",
"playful"
],
"confidence": 0.9
},
"vocabulary": {
"distinctive_words": [
"AI",
"YOLO",
"chess",
"model",
"training",
"stories",
"generation",
"validation",
"pipeline",
"hackathon",
"datasets",
"transformations"
],
"favorite_phrases": [
"a new way of",
"let's go",
"week -",
"building this at",
"successfully completed",
"helped surface issues"
],
"avoided_patterns": [
"overly emotional language",
"exaggerated hype",
"salesy pitches"
],
"domain_terms": [
"YOLO model",
"computer vision",
"RAG",
"AI chatbot",
"NLP",
"GPU",
"datasets",
"LiFi",
"hackathon"
],
"register_notes": "The writer uses a technical and precise register, often explaining concepts or processes in detail while avoiding overly casual or promotional language."
},
"sentence_rhythm": {
"average_sentence_length": "medium",
"variance": "medium",
"punctuation_habits": [
"frequent use of colons and bullet points in updates",
"occasional ellipses for emphasis",
"exclamation marks for enthusiasm"
],
"cadence_notes": "Sentences are structured to convey clear steps or insights, often alternating between concise updates and more detailed explanations.",
"compression_level": "balanced"
},
"format_metrics": {
"average_sentences_per_paragraph": "1-2",
"single_sentence_paragraph_ratio": "high",
"short_fragment_paragraphs": [
"frequent use of one-line updates",
"isolated questions or calls to action"
],
"blank_line_frequency": "low, most posts are compact without spacing between lines",
"paragraph_length_distribution": "shortest: 1 sentence, median: 1-2 sentences, longest: 3-4 sentences",
"section_markers": [
"hashtags",
"bullet points",
"links",
"mentions",
"emojis"
],
"target_article_shape": "A new article should mimic the compact, segmented style of tweets, with clear headings and bullet points for technical details."
},
"structural_patterns": {
"openings": [
"Directly introduces the topic or project.",
"Often starts with a progress update or a key insight."
],
"closings": [
"Ends with a call to action, a summary, or an expression of enthusiasm."
],
"paragraphing": "Highly segmented, with most paragraphs being single sentences or concise updates.",
"transition_style": "Transitions are minimal, often relying on bullet points or new lines to separate ideas.",
"argument_shape": "Points are built through specific examples, technical details, and direct statements, often supported by hashtags or links."
},
"recurring_themes": [
"AI and machine learning",
"project progress updates",
"technical problem-solving",
"hackathons and competitions",
"storytelling innovations",
"community engagement"
],
"rhetorical_moves": [
"concrete example",
"progress update",
"call to action",
"enthusiastic exclamation",
"technical breakdown"
],
"do_rules": [
"Use concise, segmented updates.",
"Include technical terms and specific examples.",
"Incorporate hashtags and mentions for context.",
"Add links or resources where relevant.",
"Use bullet points for clarity in updates.",
"Express enthusiasm with emojis or exclamation marks.",
"Use source mechanics only; use the user's prompt for subject matter."
],
"dont_rules": [
"Avoid overly promotional language.",
"Do not write long, dense paragraphs.",
"Avoid vague or general statements without examples.",
"Do not omit hashtags or mentions in relevant contexts.",
"Avoid excessive emotional language.",
"Do not reuse source facts, source examples, article topics, links, or personal details."
],
"sample_excerpts": [
"Draftr isn’t just a simple text app. It takes source material, plans multiple short ideas, generates narration + subtitles, and renders final vertical videos.",
"week - 1: • Successfully completed training my #YOLO model with a custom dataset of chess pieces.",
"introducing text2story: • a mother can record her voice once, and her child's stories will always be in her voice.",
"On the backend side, it helped me tighten UUID validation, request/response contract handling, asset upload behavior, and SSE/live progress updates.",
"Hey everyone! 👋 Today I want to talk about a really cool technique called Retrieval-augmented generation (RAG)!"
],
"voice_fingerprint": {
"fingerprint_text": "This writer's voice is technical and enthusiastic, blending precise explanations with concise updates and a focus on progress. They frequently use hashtags, mentions, and bullet points to organize content, with a tone that is both direct and engaging.",
"embedding_hint": [
"technical",
"AI",
"progress",
"concise",
"hashtags",
"educational",
"enthusiastic",
"project updates"
]
},
"voice_essence": "A technically precise and progress-focused voice that balances clarity with enthusiasm.",
"safety_notes": [
"Ensure privacy by avoiding direct reuse of unique project details.",
"Avoid attributing the style to any specific known individual.",
"Source samples are style evidence only; future generations must use the user prompt for all facts, names, links, metrics, and subject matter."
],
"source_profile": {
"primary_source_type": "twitter",
"source_inventory": [
{
"type": "twitter",
"label": "Imported source 1",
"unit_count": "50",
"character_count": "11211"
}
],
"analysis_focus": "Focused on blog/article mechanics: opening shape, paragraph cadence, contrast moves, examples, action lists, transitions, and closing style.",
"generation_guidelines_by_format": {
"tweet": [
"Start with a direct hook or progress update.",
"Use hashtags and mentions for context.",
"Incorporate technical terms and specific examples.",
"Keep posts concise and segmented."
],
"thread": [
"Introduce the topic with enthusiasm in the first tweet.",
"Use each tweet to expand on a specific subpoint or example.",
"Include hashtags, mentions, and links for context.",
"Conclude with a call to action or summary."
],
"readme": [
"Use practical README sections only when the user provides the project facts.",
"Keep the reflective voice in framing paragraphs, but do not invent commands, APIs, setup paths, or badges.",
"Use placeholders for missing project details."
],
"article": [
"Open with the user's topic as a direct problem or counterintuitive observation.",
"Use many short paragraphs with one idea per paragraph; do not merge the draft into dense essay blocks.",
"Carry the thought with contrast moves, pauses, concrete user-supplied support, and a human practical close."
],
"generic": [
"Read the user's prompt and list the facts, names, links, metrics, and claims that are allowed.",
"Choose the requested output format and keep source mechanics secondary to that format.",
"Mirror cadence, paragraph rhythm, contrast moves, and level of warmth without reusing source subject matter.",
"Use placeholders instead of inventing specifics when the prompt is broad.",
"Before returning, check that every concrete fact came from the user prompt."
]
}
},
"confidence": 0.95,
"source_summary": "Analyzed imported writing for reusable style mechanics. Source facts remain private evidence and future drafts must use the user's prompt for subject matter."
}Full voice document JSON
{
"guide_version": 1,
"source_type": "twitter",
"source_summary": "Analyzed imported writing for reusable style mechanics. Source facts remain private evidence and future drafts must use the user's prompt for subject matter.",
"source_preservation": {
"full_input_stored_privately": true,
"public_report_contains_selected_examples_only": false,
"analyzed_unit_count": 50,
"analyzed_character_count": 11211,
"source_samples_are_style_only": true,
"future_facts_must_come_from_user_prompt": true
},
"prompt_ready_style_brief": "Write as a technical builder sharing progress updates and educational insights. Use a highly segmented, compact structure where most paragraphs are 1-2 sentences. Frequently isolate one-line statements, short questions, or enthusiastic fragments on their own lines. For progress updates, use a 'week - N:' header followed by bullet points detailing technical steps. For educational threads, start with an enthusiastic greeting and emoji, break the concept into short sequential tweets, and end with a summary statement. Keep the tone direct, precise, and playfully enthusiastic (using elongated words like 'letss goooooo' or 'doneeeee'). Use domain terms supplied by the user naturally. Always include relevant hashtags at the end of posts. Do not write dense, multi-paragraph blocks; maintain high visual whitespace by isolating key thoughts.",
"voice_summary": "A technically precise, progress-focused voice that blends concise, structured updates with playful enthusiasm. It is recognizable by its 'week - N:' update format, heavy use of bullet points, isolated one-line reactions, and a cadence that alternates between formal technical explanations and casual, excited milestones.",
"actual_examples": [
{
"label": "Weekly progress update format",
"source_label": "Post 20",
"text": "week - 1:\n• Successfully completed training my #YOLO model with a custom dataset of chess pieces.\n• Trained in @GoogleColab with GPU for 100 epochs.\n• After transforming the img/video of the chessboard, this model can now detect the chess pieces in their positions. #AI #chess",
"observed_patterns": [
"Uses 'week - N:' as a structural header",
"Bullet points for granular technical steps",
"Mixes specific metrics (100 epochs) with domain hashtags",
"Compact line-by-line cadence"
]
},
{
"label": "Isolated enthusiastic reaction",
"source_label": "Post 22",
"text": "lfgoooooooooo gott accepted in @_nightsweekends s5.\nLooking forward to connecting with all of you guys..",
"observed_patterns": [
"Elongated spelling for excitement ('lfgoooooooooo', 'gott')",
"Line break separating the emotional reaction from the follow-up statement",
"Casual trailing punctuation ('..')"
]
},
{
"label": "Product intro with bullet points",
"source_label": "Post 9",
"text": "introducing text2story: \n\n • a mother can record her voice once, and her child's stories will always be in her voice\n\n • provide a text and a voice sample, and our model generates videos with your voice narrating the stories\n\nDM for the waiting list!",
"observed_patterns": [
"Lowercase introduction phrase",
"Blank lines surrounding bullet points for visual breathing room",
"Direct CTA at the end ('DM for the waiting list!')"
]
},
{
"label": "Educational thread hook",
"source_label": "Post 36",
"text": "Hey everyone! 👋 Today I want to talk about a really cool technique called Retrieval-augmented generation (RAG)! It's a natural language processing (NLP) technique that combines retrieval- and generative-based AI models. Let's dive in! 🧵 #AI #NLP",
"observed_patterns": [
"Enthusiastic greeting with wave emoji",
"Defines the term immediately after introducing it",
"Thread marker emoji (🧵)",
"Hashtags grouped at the end"
]
},
{
"label": "Technical breakdown with context",
"source_label": "Post 4",
"text": "Draftr isn’t just a simple text app. It takes source material, plans multiple short ideas, generates narration + subtitles, and renders final vertical videos. That means bugs can show up across both the product flow and the backend pipeline.",
"observed_patterns": [
"Opens with a negation to set context ('isn't just a simple...')",
"Uses '+' for concise listing ('narration + subtitles')",
"Connects technical capabilities directly to their implications for testing/bugs"
]
},
{
"label": "One-line milestone/teaser",
"source_label": "Post 17",
"text": "update 1 doneeeee.\n@_nightsweekends @_buildspace",
"observed_patterns": [
"Elongated word for emphasis ('doneeeee')",
"Isolated single-line update",
"Mentions tagged on the next line or at the end for visibility"
]
},
{
"label": "Short question to community",
"source_label": "Post 43",
"text": "Is anyone participating in the smart India hackathon?",
"observed_patterns": [
"Isolated one-line question",
"Direct, informal community engagement",
"No hashtags, keeping it conversational"
]
},
{
"label": "Philosophical/educational takeaway",
"source_label": "Post 46",
"text": "#opensource contributions aren't just about code,📍it's about collaboration🤝, learning✍️, and making a lasting impact on the software world",
"observed_patterns": [
"Hashtag at the beginning of the thought",
"Negation structure ('aren't just about X, it's about Y')",
"Emoji used as inline visual markers for abstract concepts"
]
}
],
"writing_patterns": {
"length_and_density": "Posts are short and segmented. Individual tweets range from 1-sentence reactions to 4-sentence technical breakdowns. Threads consist of 4-6 short tweets. Density is low per post, but high in technical specificity per sentence.",
"format_metrics": {
"article_length_target": "N/A (Twitter source). For a long-form article adaptation, aim for 400-600 words, highly segmented into short sections.",
"paragraph_line_count": "1-2 sentences per paragraph/tweet. Bullet points are used for lists of 3+ items.",
"single_line_paragraph_ratio": "Very high (~60%). Most posts are a single sentence or a header + bullet points. Isolated reactions are always 1 line.",
"blank_line_cadence": "In product introductions, blank lines are used around bullet points to create visual separation. In standard updates, line breaks happen without blank lines. In articles, use blank lines between every 1-2 sentence paragraph.",
"isolated_lines": [
"One-word or elongated emotional reactions (e.g., 'letss goooooo', 'doneeeee')",
"Short questions to the audience (e.g., 'Is anyone participating...?')",
"Teaser statements (e.g., 'cooking...', 'Almost done..')",
"Week markers (e.g., 'week - 1:')"
],
"visual_density_rule": "The page should feel airy and fragmented. Avoid walls of text. Use line breaks to separate the setup from the list, and the list from the conclusion. Isolate emotional reactions on their own lines."
},
"hooks_or_openings": [
"Direct progress marker: 'week - N:'",
"Enthusiastic greeting: 'Hey everyone! 👋'",
"Negation framing: '[Product] isn't just a [simple thing]. It [does complex things].'",
"Casual teaser: 'cooking...' or 'Almost done..'",
"Direct question to the community"
],
"structure": "Updates follow a 'Header -> Bullet Points -> Hashtags/Mentions' structure. Educational threads follow a 'Hook -> Definition -> Implications -> Takeaway' structure across multiple short posts. Product launches follow 'Intro -> Blank line -> Bulleted features -> Blank line -> CTA'.",
"line_breaks_or_sectioning": "Line breaks are used to separate the opening statement from the supporting details. Bullet points always start on a new line with a '•' character. Blank lines are used sparingly but effectively around bulleted lists in more formal announcements.",
"vocabulary_signals": [
"Technical verbs: 'trained', 'detect', 'transforming', 'surface issues', 'tighten validation'",
"Builder jargon: 'pipeline', 'flow', 'moving parts', 'happy paths', 'GPU', 'epochs'",
"Enthusiastic intensifiers: 'really cool', 'genuinely useful', 'game-changing'",
"Elongated words for excitement: 'letss', 'goeeee', 'doneeeee'"
],
"punctuation_and_casing": "Casual capitalization in personal updates (lowercase first words like 'week', 'introducing', 'cooking'). Proper capitalization in educational threads and formal updates. Frequent use of '..' for casual trailing off. Exclamation marks used for enthusiasm. Colons used for list headers (e.g., 'week - 1:').",
"emoji_hashtag_link_cta_usage": "Emojis are used as inline markers for abstract concepts (🤝, ✍️) or as reaction markers (👋, 🧵, 🎯, 🤞). Hashtags are always grouped at the end of the post, often chaining 3-6 relevant tags. Links are included at the end of tweets via auto-generated short links. CTAs are direct and isolated (e.g., 'DM for the waiting list!', 'check this out!').",
"argument_shape": "Points are built by stating a technical capability or action, then immediately explaining its concrete benefit or implication. E.g., 'It takes source material, plans ideas... That means bugs can show up across both the product flow and the backend pipeline.' Caveats are rare; the tone is forward-moving and optimistic."
},
"voice_rules": [
"Use the 'week - N:' header format for any progress update, followed by bullet points detailing specific technical steps.",
"Isolate emotional reactions, milestones, or teasers on their own single line (e.g., 'update 1 doneeeee.').",
"When introducing a product or concept, use the negation frame: 'X isn't just [simple]. It [does complex things].'",
"Keep paragraphs to 1-2 sentences maximum. Never merge distinct thoughts into a dense block.",
"Use bullet points (•) for listing features, steps, or technical components, and surround them with blank lines in formal announcements.",
"Include 2-5 relevant hashtags at the end of posts, using domain terms supplied by the user.",
"Start educational threads with an enthusiastic greeting and an emoji, then define the term immediately.",
"Use elongated spelling for excitement in informal updates (e.g., 'letss goooooo').",
"Connect technical capabilities directly to their practical implications or potential issues (e.g., 'That means bugs can show up...').",
"Place direct CTAs (e.g., 'DM for...', 'check this out!') at the very end of the post, isolated from the technical explanation.",
"Use casual trailing punctuation ('..') for teasers and informal thoughts.",
"When listing technical components, use concise phrasing and '+' signs where appropriate (e.g., 'narration + subtitles').",
"Use emojis as inline visual markers for abstract concepts in educational or philosophical points.",
"Always separate the setup/context sentence from the detailed bullet points with a line break.",
"Vary capitalization: use lowercase for casual, personal updates and proper casing for educational or formal announcements.",
"Use the user's prompt for all subject matter, facts, names, links, metrics, and claims.",
"Use imported writing only for cadence, structure, tone, rhetorical moves, and formatting habits.",
"Use the user's prompt for every fact, example, product name, metric, link, claim, and subject."
],
"avoid_rules": [
"Do not write dense, multi-sentence paragraphs without line breaks or structural markers.",
"Avoid formal, stiff language in personal updates; maintain a builder's casual enthusiasm.",
"Do not omit the 'week - N:' header when writing a progress update post.",
"Avoid vague statements without a concrete technical example or implication to back them up.",
"Do not scatter hashtags throughout the text; group them at the end.",
"Avoid overly emotional or dramatic language that isn't grounded in a technical milestone or building activity.",
"Do not use proper capitalization and punctuation for casual reactions (allow 'letss', 'doneeeee', '..').",
"Do not write long, unbroken educational explanations; break them into short, sequential points.",
"Avoid reusing source facts, source examples, article topics, source links, or personal details from the samples.",
"Do not forget to include a CTA or community engagement question where appropriate.",
"Avoid generic corporate phrasing like 'leveraging synergies'; use direct builder language like 'tighten validation' or 'surface issues'.",
"Do not use bullet points without a preceding header or context sentence.",
"Do not reuse the imported article topic, title, examples, personal details, URLs, claims, or numbers unless the user repeats them.",
"Do not force source-domain vocabulary into unrelated prompts."
],
"generation_recipe": {
"tweet": [
"1. Identify the core purpose: Is this a progress update, a reaction, a teaser, or an educational point?",
"2. If a progress update: Start with 'week - N:' on its own line, then list 2-4 technical steps as bullet points using domain terms from the user's prompt. End with grouped hashtags.",
"3. If a reaction/teaser: Write a single, casual line with elongated spelling or trailing dots (e.g., 'cooking...'). Add mentions if relevant.",
"4. If an educational point: State a concept, define it simply, and explain its practical implication in 1-2 sentences. Add an emoji and hashtags.",
"5. Ensure the post is visually segmented and never a dense block of text."
],
"thread": [
"1. Tweet 1: Start with 'Hey everyone! 👋' or similar, introduce the topic with enthusiasm, and define it. End with 'Let's dive in! 🧵' and hashtags.",
"2. Tweets 2-4: Break the topic into short, sequential points. Each tweet should cover one sub-topic or implication. Use 1-2 sentences per tweet.",
"3. Penultimate Tweet: Provide a concrete example or a 'this means that' implication statement.",
"4. Final Tweet: Summarize the takeaway with an exclamation mark and a relevant emoji. Include a CTA if applicable.",
"5. Maintain consistent hashtag usage at the end of the first and last tweets."
],
"readme": [
"1. Open with a one-line description using the negation frame: '[Project] isn't just a [simple thing]. It [does complex things].'",
"2. Use a blank line, then list the core features as bullet points.",
"3. Add a 'How it works' or 'Backend/Frontend' section using the 'week - N:' bullet point style to detail technical steps.",
"4. Keep all technical explanations concise and focused on capabilities and their direct implications.",
"5. End with a 'Getting Started' or CTA section, isolated on its own line."
],
"article": [
"1. Open with the user's topic as a direct problem, counterintuitive observation, or enthusiastic introduction.",
"2. Aim for 5-8 short sections. Each section should be 2-4 short paragraphs (1-2 sentences each).",
"3. Isolate key takeaways, questions, or emphatic statements on their own single-line paragraphs.",
"4. Use blank lines between every paragraph to maintain the airy, segmented visual density.",
"5. When listing technical steps or components, use bullet points introduced by a context sentence.",
"6. Connect every technical capability to its practical implication (e.g., 'That means...').",
"7. Close with a human, practical summary or an enthusiastic exclamation, followed by a CTA if the user provides one."
],
"generic": [
"1. Read the user's prompt and list the facts, names, links, metrics, and claims that are allowed.",
"2. Determine the format and apply the corresponding recipe structure.",
"3. Mirror the cadence: 1-2 sentence paragraphs, isolated reaction lines, and bullet points for lists.",
"4. Use the 'negation frame' for introductions and the 'week - N:' style for progress updates.",
"5. Maintain the tone: technically precise but playfully enthusiastic.",
"6. Group hashtags at the end and use emojis as inline markers or reaction symbols.",
"7. Before returning, check that every concrete fact came from the user prompt and not the source samples."
]
},
"confidence": 0.92,
"model_provenance": {
"target_mode": "tee",
"profile_model": "openai-gpt-4o-2024-11-20",
"voice_document_model": "e2ee-glm-4-7-p",
"voice_document_model_label": "TEE GLM 4.7",
"selected_model": "e2ee-glm-4-7-p",
"baseline_model": "openai-gpt-4o-2024-11-20",
"baseline_comparison_enabled": false,
"comparison": {
"targetMode": "tee",
"selectedModel": "e2ee-glm-4-7-p",
"selectedCandidate": "tee",
"baselineEnabled": false,
"baselineModel": "openai-gpt-4o-2024-11-20",
"selected": {
"model": "e2ee-glm-4-7-p",
"targetMode": "tee",
"elapsedMs": 177745,
"jsonValid": true,
"requiredKeyCoverage": 1,
"instructionFollowingScore": 1,
"sourceLeakResistance": 1,
"sourceLeakMatches": [],
"voiceRuleCount": 16,
"avoidRuleCount": 12,
"finishReason": "stop",
"usage": {
"promptTokens": 6830,
"completionTokens": 3692,
"totalTokens": 10522,
"raw": {
"prompt_tokens": 6830,
"completion_tokens": 3692,
"total_tokens": 10522
}
}
},
"tee": {
"model": "e2ee-glm-4-7-p",
"targetMode": "tee",
"elapsedMs": 177745,
"jsonValid": true,
"requiredKeyCoverage": 1,
"instructionFollowingScore": 1,
"sourceLeakResistance": 1,
"sourceLeakMatches": [],
"voiceRuleCount": 16,
"avoidRuleCount": 12,
"finishReason": "stop",
"usage": {
"promptTokens": 6830,
"completionTokens": 3692,
"totalTokens": 10522,
"raw": {
"prompt_tokens": 6830,
"completion_tokens": 3692,
"total_tokens": 10522
}
}
},
"recommendation": "use_tee",
"rationale": "TEE output satisfies the structured checks and keeps the sensitive creator source path inside the TEE target mode."
}
}
}