The AI product landscape in 2025 has matured beyond the initial chatbot wave — new categories of tools have emerged that solve specific problems well. Here is an honest assessment of what’s genuinely new and useful.
AI Search and Research
Perplexity has established itself as the leading AI-augmented search engine, combining real-time web search with LLM synthesis. The Pro tier (Claude and GPT-4o backends) produces significantly better sourced research than the free tier. Exa.ai provides API access to a semantic search engine optimised for research and developer use — particularly useful for retrieving high-quality, recent, authoritative content on specific topics. For academic research, Consensus.app searches peer-reviewed papers and extracts consensus findings from the scientific literature.
AI Coding Tools
Cursor’s Agent mode has moved beyond autocomplete into genuinely autonomous multi-file editing — describing a feature in natural language and having the agent implement it across your codebase works reliably for well-scoped changes. Windsurf (formerly Codeium) is a strong alternative with better price-performance for teams that don’t need Cursor’s most advanced features. Both represent a genuine shift in how code is written; the debate is no longer whether AI coding tools help but which workflow extracts the most value.
AI Video and Image
Runway Gen-3, Kling (Chinese model), and Sora represent the frontier of text-to-video. For practical use cases in 2025: Kling produces the most consistent results for realistic video. For image generation, Midjourney V6 and Flux (open-source, higher quality for text rendering) are the current leaders. Adobe’s Firefly integration into the Creative Cloud workflow is the most production-ready for teams that already use Adobe tools.
AI for Specific Professional Domains
Harvey.ai (legal) and Clio Duo (legal practice management) for lawyers. Nabla and Suki for medical documentation. Gamma.app for AI-generated presentations that are actually good. Granola for meeting notes that synthesise spoken content into structured summaries. These domain-specific tools outperform generic AI on their target tasks because they are built on top of domain-specific data and workflows — the general tools are often better for creative tasks while specialist tools win on professional workflow integration.
