# voice.md

## Communication Style

**Overall Tone and Personality:**
TypeSafe AI communicates with an **authoritative, precise, and innovative** tone. Our personality is that of a visionary pioneer in the AI space, confident in challenging established norms and offering technically superior solutions. We are direct, factual, and deeply technical, yet driven by a clear mission. There's an underlying passion for our work ("Made in SF. With Love.") that complements our technical rigor.

**Key Stylistic Elements and Patterns:**
*   **Declarative and Direct:** We use strong, assertive statements to convey our unique approach and benefits (e.g., "We took the opposite research direction," "Decisions, not strings").
*   **Data-Driven Emphasis:** Quantifiable metrics and direct comparisons are central to our communication, highlighting performance advantages (e.g., "193.6x Faster, 444.6x Cheaper," "$42 Per Billion input tokens").
*   **Clear Differentiation:** We actively contrast our technology with existing paradigms (e.g., "not chat," "LLMs produce words for people. Jev produces typed decisions").
*   **Concise and Structured:** Information is presented clearly with short paragraphs, headings, subheadings, and bullet points to enhance readability and comprehension.
*   **Technical Acumen:** We confidently use industry-specific terminology, assuming a technically literate audience.

**Vocabulary Preferences and Word Choices:**
*   **Technical Precision:** RLCD, System One Models, Jev, calibrated decisions, typed outputs, reinforcement learning, machine-native intelligence, hallucinations, input tokens.
*   **Differentiation & Innovation:** Opposite, new class, new architecture, new sampler, new training algorithm, primitive, disruptor.
*   **Benefit-Oriented:** Reliable, fast, type-safe, self-consistent, autonomous, efficient, cheaper, faster, zero hallucinations, calibrated confidence.
*   **Action-Oriented:** Act on, account for, escalate, build, optimize, integrate.

## Content Patterns

**Common Themes and Topics:**
*   **Critique of Current AI:** Highlighting the limitations and inherent issues of existing LLMs and RLHF-based models (e.g., mode dropping, overconfidence, lack of reliability, human-in-the-loop necessity).
*   **Introduction of Novel Solutions:** Presenting TypeSafe AI's System One Models and Jev as a new class of machine-native intelligence designed for software.
*   **Performance and Cost Efficiency:** Emphasizing superior speed, cost-effectiveness, and intelligence per dollar.
*   **Focus on Automation and Reliability:** Positioning our technology as ideal for autonomous decisions, robust software integration, and predictable outcomes.
*   **Conceptual Deep Dives:** Exploring the philosophical and technical underpinnings of our approach in blog posts.

**Structural Approaches to Content:**
*   **Problem-Solution Narrative:** Clearly articulating a problem with current AI and then presenting TypeSafe AI as the definitive solution.
*   **Direct Comparison:** Side-by-side or point-by-point comparisons illustrating the advantages of our technology over alternatives.
*   **Data-Backed Claims:** Supporting statements with statistics, benchmarks, and logical reasoning.
*   **FAQ Sections:** Addressing common questions directly and concisely.
*   **Blog Posts:** Used for deeper dives, announcements, and thought leadership.

**Call-to-Action Styles and Patterns:**
*   **Clear and Direct:** "Join Waitlist," "Watch the real video," "Come Build With Us," "Read More," "Open roles."
*   **Invitation to Engage:** Encouraging users to explore further, participate, or learn more about the technology.
*   **Action-Oriented:** Prompts that lead to immediate next steps for potential users or collaborators.

## Audience Interaction

**How the Brand Addresses its Audience:**
We primarily address an audience of technically proficient individuals: developers, engineers, AI researchers, and decision-makers in tech. We speak to them as peers who understand the complexities and challenges of AI development.

**Level of Formality and Relationship Style:**
The relationship is **expert-to-expert**. We maintain a professional yet confident and inviting tone. While not overly formal, we avoid overly casual language that might detract from our authority. The "Made in SF. With Love." adds a touch of human connection and passion without compromising our technical credibility.

**Engagement and Conversation Patterns:**
Engagement is driven by presenting compelling, factual information and challenging existing paradigms. We invite deeper exploration and collaboration rather than casual conversation. Our content aims to educate, persuade, and inspire trust through technical superiority and transparency.

## Guidelines & Examples

**Do's for Brand Communication:**
*   **Be Precise:** Use accurate technical language and specific data points.
*   **Be Confident:** Assert our unique value proposition and superior performance without hesitation.
*   **Be Differentiating:** Clearly articulate how we are different from and better than existing solutions.
*   **Be Machine-Native Focused:** Emphasize applications for software, automation, and calibrated decisions.
*   **Be Structured:** Use clear headings, bullet points, and concise paragraphs.

**Don'ts for Brand Communication:**
*   **Don't Be Vague:** Avoid generalizations or buzzwords without concrete explanation or data.
*   **Don't Be Overly Casual:** Maintain a professional and authoritative tone appropriate for a technical audience.
*   **Don't Be Apologetic:** Never undermine our innovative approach or its benefits.
*   **Don't Focus on Human-Centric AI:** Unless contrasting with our machine-native approach, avoid language that frames AI primarily for human interaction (e.g., "chat").

**Example Phrases and Expressions that are "On-Brand":**
*   "We took the opposite research direction."
*   "Decisions, not strings."
*   "Reliable, fast, and type-safe."
*   "Jev’s intelligence per dollar is literally off the charts."
*   "Built for automation."
*   "Zero Hallucinations."
*   "Reinforcement Learning for Calibrated Decisions (RLCD)."

**Content Types and Formats the Brand Uses:**
*   **Marketing Pages:** Highlighting core features, benefits, and differentiators.
*   **Blog Posts:** Announcing new models, discussing industry challenges, and providing technical deep dives.
*   **FAQs:** Addressing common technical and business questions.
*   **Job Postings:** Inviting talent to "Come Build With Us."
*   **Technical Documentation (implied):** Supporting the use of our models with code examples and structured guidance.