Liminary Competition
By Mitchell Hart
About this collection
## AI Tools for Human Augmentation: Sentience & TypeOS This two-document collection covers two distinct AI products that share a common thread: **using AI to extend human cognitive capability** rather than replace it. - **Sentience** is building a "digital version of you" — a 7-phase roadmap to emulate the human mind (perception → memory → recall → attention → motivation → emotion → consciousness), currently ranging from 21–75% complete across phases. - **TypeOS** is a Google Docs AI writing assistant that emphasizes human *taste* and control — offering editing, research, AI detection bypass, and style-matching features, trusted by 1,000+ users. Both products position AI as a layer that amplifies human judgment rather than replacing it. Sentience takes a neuroscience-grounded systems approach; TypeOS takes a practical, workflow-integrated approach. Together they represent two ends of the AI augmentation spectrum: one deeply theoretical and long-horizon, the other immediately deployable and commercially focused. Relevant to anyone researching the AI tools landscape, knowledge management, or competitive positioning in the writing/productivity space.
Curated Sources
Liminary - Storage for the AI Age
Liminary is an AI-native storage platform designed to help users efficiently capture, organize, and recall knowledge from various sources. It enables one-click saving of content from webpages, PDFs, YouTube videos, AI chats, and emails, eliminating the need for manual searching or tab switching. The platform surfaces relevant information while users are reading, writing, or researching, and automatically connects old ideas with new ones to enrich knowledge over time. Liminary is targeted at professionals who need to synthesize large amounts of information quickly, such as consultants, researchers, venture capitalists, and content creators, offering a seamless way to retrieve and utilize past insights without disrupting their thinking process.
Key Takeaways
- Liminary provides a unified solution for saving and recalling information from diverse sources like webpages, PDFs, videos, and AI chats, eliminating fragmented workflows.
- It surfaces relevant knowledge contextually while users work, reducing time spent searching for past insights and improving decision-making efficiency.
- The platform is designed for professionals who need to synthesize large amounts of information quickly, such as consultants, researchers, and venture capitalists, by connecting old and new ideas automatically.
Sentience | Build the digital version of you
The Sentience Company outlines a comprehensive seven-phase roadmap to emulate human cognition through software systems that mirror biological brain functions. Rejecting connectomics approaches, the plan focuses on creating high-fidelity analogs of human mental systems using current and emerging technologies. Phase 1 (Perception) captures multi-modal sensory data (vision, audio, text) with 75% completion. Phase 2 (Memory) implements working, episodic, semantic, and procedural memory systems with 75% completion. Phase 3 (Recall & Retrieval) enables cue-driven reconstruction of information at 50% completion. Phase 4 (Attention & Executive Function) develops cognitive control mechanisms at 36% completion. Phases 5-6 (Motivation & Emotional Architecture) represent research stages at 25-33% completion. The final Phase 7 tackles consciousness and full emulation through metacognition and global workspace modeling at 21% completion. The approach integrates neuroscience frameworks like Baddeley's working memory model, ACT-R spreading activation, and Tononi's Integrated Information Theory with modern AI techniques including vector embeddings, RAG agents, and LLM-based processing.
Key Takeaways
- The company prioritizes mirroring human cognitive systems over connectomics, using software analogs of biological structures like hippocampus and prefrontal cortex
- Current implementation focuses on perception and memory systems with 75% completion, while attention mechanisms are at 36% and consciousness research at 21%
- The roadmap integrates established neuroscience models (Baddeley, ACT-R, IIT) with modern AI techniques for practical implementation
- Progress tracking shows concrete technical achievements in data capture and memory storage, with later phases remaining research-focused
- The approach acknowledges fundamental challenges in consciousness research while pursuing incremental technical implementation
TypeOS | AI writing you control with your taste.
TypeOS is an AI-powered writing assistant seamlessly integrated into Google Docs, designed to enhance productivity for students, researchers, and professionals. The tool offers real-time editing suggestions that users can accept or reject, enabling granular control over AI-generated content. It features specialized capabilities like humanizing AI text to bypass major detection systems (GPTZero, Turnitin, Originality.ai), instant research summarization with source integration, and automated formatting for academic styles (APA, MLA, Chicago). TypeOS supports multiple cutting-edge AI models including GPT-4, Claude, and Gemini, adapting to user writing styles through context-aware processing. The platform includes research tools like preference learning analysis, TypeBench for measuring human taste in writing, and FormatBench for evaluating structural fidelity across document types. Testimonials highlight its effectiveness for last-minute academic papers, professional documents, and creative writing, with users praising its ability to handle references, improve clarity, and maintain original voice. The service addresses key pain points in modern writing workflows by combining advanced AI capabilities with native Google Docs compatibility, positioning itself as a productivity layer that preserves human judgment while automating repetitive tasks.
Key Takeaways
- TypeOS integrates directly into Google Docs to provide context-aware AI editing without switching tools, increasing adoption and workflow efficiency.
- The tool's humanization feature effectively bypasses major AI detection systems, making it valuable for academic and professional writing where authenticity is critical.
- Underpinning research on preference learning and writing style adaptation could influence future AI writing models by capturing nuanced human taste through micro-editing decisions.
- Support for multiple AI models and automated formatting addresses diverse use cases from academic papers to business contracts in a single interface.
- Real-world adoption by students and professionals demonstrates practical value for time-sensitive document creation and quality improvement.
Frequently Asked Questions
- How does Sentience's memory architecture (PostgreSQL + TurboPuffer vectors) compare to how tools like Recall or Rewind handle episodic memory storage?
- TypeOS claims to pass GPTZero, Turnitin, and Originality.ai at 0% AI — what are the ethical and legal risks of this 'Humanize' feature for their enterprise/institutional market?
- Sentience is at 75% on Perception and Memory but only 36% on Attention/Executive Function — does this mean it can capture information well but can't yet act on it intelligently?
- Both Sentience and TypeOS position human 'taste' or judgment as the irreplaceable layer — how does this framing hold up as AI models get more capable at mimicking taste?
- TypeOS's research page mentions a 'writing preference dataset' built from accept/reject decisions — could this data moat be their real competitive advantage over ChatGPT or Gemini in Docs?
- Sentience's Phase 7 (Consciousness) cites IIT and Tononi's Φ — do they believe their feed-forward architecture can actually achieve high Φ, or is consciousness just aspirational branding?
- How do Sentience and TypeOS each handle privacy — and which approach is more credible for enterprise or academic adoption?