Bonkers

Generates original written content (articles, blog posts, emails, social media copy) by routing user prompts through OpenAI's GPT-4 API with context-aware instruction templates. The system maintains conversation history within browser sessions to enable iterative refinement, allowing users to request rewrites, tone adjustments, or expansions without re-specifying the full context. Integration with browser extension allows in-context generation directly within web applications (Gmail, Google Docs, etc.) by capturing surrounding text as implicit context.

Accepts long-form text (articles, PDFs, emails, research papers) and generates concise summaries using GPT-4 with configurable output length (bullet points, paragraph, or key takeaways). The system uses prompt engineering to enforce summary constraints rather than token-limiting, allowing users to specify desired granularity (executive summary vs. detailed outline). Browser extension can auto-summarize web articles on demand by extracting main content via DOM parsing.

Generates code snippets, functions, and full scripts across multiple programming languages (Python, JavaScript, Java, C++, etc.) by accepting natural language descriptions or partial code and returning complete, executable implementations. Uses GPT-4's code understanding to infer intent from context (e.g., 'sort this array' generates language-appropriate sorting logic). Browser extension allows in-context code generation within code editors (VS Code, GitHub, etc.) by capturing surrounding code as implicit context for coherent suggestions.

Analyzes written text for grammatical errors, punctuation issues, and stylistic improvements, then provides corrected versions with optional tone adjustments (formal, casual, persuasive, etc.). Uses GPT-4's language understanding to preserve original meaning while enhancing clarity and readability. Browser extension integrates with web-based text editors (Gmail, Google Docs, LinkedIn, etc.) to offer in-place corrections without copying text out of context.

Generates images from natural language prompts by routing descriptions through an image generation API (likely DALL-E or similar) integrated with Merlin's backend. Users provide text descriptions of desired images, and the system returns generated images in standard formats (PNG, JPEG). Quality and style control depend on prompt engineering and underlying model capabilities.

Deploys a browser extension that injects AI assistance into web-based applications (Gmail, Google Docs, LinkedIn, GitHub, etc.) by capturing surrounding text/code as implicit context and offering relevant suggestions without tab-switching. The extension maintains a persistent UI panel for accessing Merlin's capabilities (writing, summarization, code generation) while staying within the current application. Context capture uses DOM parsing to extract relevant content and pass it to GPT-4 for contextually-aware responses.

Provides free-tier access to GPT-4 capabilities with limited monthly usage (exact limits unknown), and paid tiers for higher usage. The freemium model routes user requests through Merlin's backend API, which abstracts OpenAI's GPT-4 API and applies rate limiting and quota management. Users can upgrade to paid tiers for increased token limits and priority processing. Pricing structure and tier details are not transparently documented.

Maintains conversation history within browser extension sessions, allowing users to reference previous messages and build on prior responses without re-specifying full context. Each conversation thread preserves the full exchange with GPT-4, enabling iterative refinement (e.g., 'make it shorter', 'add more examples', 'change the tone'). Context is stored locally in browser storage or session memory; persistence across browser restarts is unknown.