storm vs tldraw Make Real

Generates research questions through simulated conversations between a Wikipedia writer and topic expert LLM agents, where questions are grounded in perspective discovery from similar existing articles rather than direct prompting. The system surveys related Wikipedia articles to extract diverse viewpoints, then uses these perspectives to guide the question-asking process, ensuring comprehensive topic coverage from multiple angles. This two-agent conversational approach with perspective injection produces more structured and comprehensive research directions than naive question generation.

Unique: Uses perspective discovery from existing articles to guide question generation rather than direct LLM prompting, implemented as a two-agent conversation (Wikipedia writer + topic expert) that grounds questions in retrieved reference patterns. This contrasts with naive question generation that lacks structural guidance from domain knowledge organization.

vs alternatives: Produces more comprehensive and well-organized research questions than single-prompt approaches because it learns perspective structure from authoritative sources rather than relying on LLM priors alone.

Generates multi-level article outlines (sections, subsections, key points) using collected research references, where each outline node is anchored to specific retrieved sources. The system structures the outline hierarchically to match Wikipedia article conventions, then maps each outline element to supporting citations from the knowledge curation phase. This enables the subsequent writing stage to generate text with proper in-line citations by maintaining explicit outline-to-source mappings throughout the generation pipeline.

Unique: Maintains explicit outline-to-source mappings throughout generation, enabling downstream article writing to produce citations without additional retrieval. The outline generation phase explicitly anchors each structural element to supporting references from the knowledge curation phase, creating a citation-aware outline rather than a generic structure.

vs alternatives: Guarantees citation availability at write time because outline generation is citation-aware, whereas generic outline generators may create structures that lack source support.

Orchestrates the complete STORM pipeline (knowledge curation → outline generation → article writing → polishing) for batch processing of multiple topics, implemented through STORMWikiRunner that manages state, error handling, and progress tracking across pipeline stages. The system executes each stage sequentially for each topic, maintaining intermediate results and enabling resumption from failure points. This orchestration layer abstracts pipeline complexity and enables users to generate article collections without managing individual stage invocations.

Unique: Implements STORMWikiRunner that orchestrates the complete multi-stage pipeline (knowledge curation → outline → article → polish) with state management and error handling, enabling batch article generation without manual stage invocation. The runner maintains intermediate results and enables resumption from failure points.

vs alternatives: Simplifies batch article generation compared to manual stage invocation because the runner handles pipeline orchestration, state management, and error handling transparently.

Uses sentence encoders (embeddings) to compute semantic similarity between research questions and existing article content, enabling the system to discover relevant perspectives from similar articles without explicit keyword matching. The encoder system converts text to dense vector representations, enabling efficient similarity search across large article collections. This semantic approach discovers perspectives that keyword-based methods would miss, improving the diversity and relevance of research questions.

Unique: Uses sentence encoders to compute semantic similarity for perspective discovery, enabling the system to find relevant perspectives from similar articles based on meaning rather than keywords. This semantic approach discovers diverse perspectives that keyword matching would miss.

vs alternatives: Discovers more diverse and relevant perspectives than keyword-based methods because semantic similarity captures meaning-level relationships rather than surface-level term overlap.

Generates full-length Wikipedia-style articles (2000+ words) by consuming hierarchical outlines and mapped citations, producing text with inline citations that reference specific retrieved sources. The system uses the outline structure to guide section-by-section generation, maintaining citation context from the outline-to-source mappings to ensure every claim references a specific source. This multi-stage approach (outline → section generation → citation insertion) produces coherent long-form content with proper attribution without requiring additional source retrieval during writing.

Unique: Generates long-form articles with inline citations by leveraging pre-computed outline-to-source mappings from the outline generation phase, eliminating the need for citation lookup during writing. The system maintains citation context throughout multi-section generation, enabling coherent long-form text with proper attribution without additional retrieval.

vs alternatives: Produces properly cited long-form content more efficiently than retrieval-augmented generation approaches that re-fetch sources during writing, because citation mappings are pre-computed in the outline phase.

Integrates with internet search APIs (Bing, Google, or custom) to retrieve relevant sources for research questions, implementing a retrieval module that handles query expansion, result ranking, and content extraction. The system executes search queries derived from research questions, collects results with metadata (URLs, snippets, relevance scores), and extracts full-text content from retrieved pages. This retrieval layer feeds the knowledge curation phase with grounded source material, enabling all downstream stages to operate on internet-sourced information.

Unique: Implements a pluggable retrieval module that abstracts search provider (Bing, Google, custom) and handles full-text extraction from retrieved pages, enabling the knowledge curation pipeline to operate on rich source content rather than search snippets alone. The retrieval layer maintains source metadata throughout the pipeline for citation purposes.

vs alternatives: Provides richer source material than snippet-only search because it extracts full-text content from retrieved pages, enabling more comprehensive knowledge curation and citation accuracy.

Builds and maintains a hierarchical knowledge base (mind map) that organizes collected information into a dynamic concept structure, implemented as the KnowledgeBase class that stores information as nested concepts with relationships. The system continuously reorganizes information as new sources are added, maintaining a shared conceptual space that reduces cognitive load during knowledge curation. This knowledge base serves as the source of truth for outline generation and article writing, enabling both automated and human-collaborative workflows to reference a consistent information structure.

Unique: Maintains a dynamic, reorganizable knowledge base that serves as a shared reference structure for both automated and human-collaborative workflows, implemented as a hierarchical concept map that evolves as new information is added. This contrasts with static information tables that don't reorganize or provide cognitive scaffolding for long research sessions.

vs alternatives: Enables human-AI collaborative research more effectively than flat information tables because the hierarchical concept structure provides cognitive scaffolding and reduces information overload during extended curation sessions.

Implements a three-agent collaborative discourse protocol (Co-STORM) where human users, LLM expert agents, and a moderator agent participate in structured knowledge curation conversations. The moderator agent generates thought-provoking questions inspired by retrieved information not yet discussed, expert agents answer questions grounded in external sources and raise follow-up questions, and human users can observe passively or actively steer the conversation. The system maintains conversation history and the shared knowledge base, enabling the moderator to track discussed vs. undiscussed information and guide the discourse toward comprehensive coverage.

Unique: Implements a three-agent collaborative protocol with explicit moderator coordination that tracks discussed vs. undiscussed information and generates targeted follow-up questions, enabling human-AI research teams to maintain conversation coherence and comprehensive coverage. The moderator agent explicitly inspects the knowledge base to identify information gaps and guide the discourse.

vs alternatives: Enables more comprehensive and coherent human-AI collaboration than simple chatbot interfaces because the moderator agent actively tracks coverage and generates targeted follow-up questions rather than passively responding to user input.

This capability allows users to convert hand-drawn UI sketches into functional HTML, CSS, and JavaScript code by leveraging AI vision. When a user clicks the 'Make it Real' button, the system captures the drawn elements and sends them to an AI provider (like OpenAI or Anthropic) via a Next.js API route. The AI processes the input and returns structured code, which is then displayed in a preview component. This integration with multiple AI providers enables flexibility in the transformation process.

Unique: Utilizes a custom hook (useMakeReal) to orchestrate the transformation process, managing state and API interactions seamlessly.

vs alternatives: More intuitive than traditional design-to-code tools, as it directly interprets hand-drawn inputs.

Make Real integrates with multiple AI providers through dynamic Next.js API routes, allowing users to select their preferred AI service for code generation. The application uses a modular architecture where each provider's API is handled separately, enabling easy updates and maintenance. This design allows for a seamless user experience as the system can switch between providers based on user settings without altering the core functionality.

Unique: Supports multiple AI providers through a single interface, allowing easy switching and configuration via a settings dialog.

vs alternatives: More adaptable than single-provider solutions, providing users with options based on their needs.

This capability provides users with a real-time preview of the generated HTML output within the application. After the AI processes the sketch, the resulting HTML is rendered in a dedicated preview component (PreviewShape). This allows users to see immediate feedback on their designs, facilitating rapid iterations and adjustments. The use of React components ensures that the UI remains responsive and interactive during the preview process.

Unique: Integrates a dedicated preview component that updates dynamically as users modify their sketches, enhancing the prototyping experience.

vs alternatives: Offers a more interactive and immediate feedback loop compared to traditional design tools that require separate preview steps.

The system extracts textual information from the selected shapes in the Tldraw editor using a custom utility. This process involves analyzing the drawn elements to identify and capture any text, which is then formatted into a prompt for the AI provider. This capability is crucial for generating accurate code, as it ensures that all relevant information from the sketches is utilized during the transformation process.

Unique: Employs a specialized text extraction utility that focuses on shapes within the Tldraw canvas, enhancing the accuracy of the generated prompts.

vs alternatives: More tailored for sketch-based inputs than generic OCR tools, providing context-aware text extraction.

Make Real includes a settings management system that allows users to configure their preferences, such as selecting AI providers and entering API keys. This functionality is managed through a dedicated settings dialog that persists user configurations in localStorage. The design ensures that user preferences are retained across sessions, enhancing the overall user experience and making it easy to switch between different setups.

Unique: Utilizes localStorage to persist user settings, allowing for quick retrieval and modification without server-side dependencies.

vs alternatives: More user-friendly than manual configuration files, as it provides a straightforward UI for managing settings.

tldraw Make Real is an innovative web application that transforms hand-drawn sketches and wireframes into functional HTML/CSS/JavaScript code using AI vision, making it ideal for rapid prototyping and turning napkin sketches into working UIs.

Unique: This tool uniquely combines hand-drawn input with AI to generate working code, streamlining the design-to-development process.

vs alternatives: Unlike traditional coding tools, tldraw Make Real allows users to visually create interfaces and instantly convert them to code, significantly speeding up the prototyping phase.