Korva is a scroll of the internet’s best ideas.
Model makers are an asset
Everything Anthropic, OpenAI, Google, and open source are building is good news for Glean, which could not have delivered its product without that help. Founders who lose sleep over being eaten should focus on solving problems first, while still watching what these companies can do today.
Anthropic's vertical packs for areas like design, legal, and finance are quite shallow, and few people, if any, move their workload entirely from Figma or another tool to Anthropic. The use is net new. Designers still use Figma, while non-designers use Claude's design tool, because AI makes things simpler for people who are not experts in a tool.
Glean meets this competition every day. Customers ask what it does that Claude, connected to enterprise systems through MCP, cannot. Question answering is the largest use of AI in the world, and Claude Cowork and Claude Desktop have always been built around it. Being first to bring RAG and semantic search to the enterprise gave Glean its brand and the right to compete, but it is neither a requirement nor a savior.
How did RAG become standard for enterprise AI?
Retrieval-augmented generation was described in a 2020 paper by Patrick Lewis and colleagues at Facebook AI Research. A model only knows what was in its training data, and private company documents were never in it. RAG fixes this by searching a company's own documents at question time and handing the relevant passages to the model. It also lets answers point back to specific sources.
Why do design tools draw non-expert users?
Professional design software has many menus and a steep learning curve built for people who use it every day. Canva launched in 2013 aimed at people with no design training, using templates and simple drag-and-drop editing. Chat-based tools lower the bar further, because users describe what they want in words. That draws people who would never have opened a professional tool, which makes much of the use new instead of moved.