“People wanted to see more pro-code flexibility,” he said in an interview at Sapphire 2026. “We had gone with a low-code approach. You could give it extension points and tools, but you couldn't touch the core of it. Now you can build a custom agent, connect it to your own GitHub.”
Glean Chat offers an experience very similar to OpenAI's ChatGPT, but limited to an enterprise's content and resource boundaries, Jain said. When a user makes a natural language-based query, the company's search technology uses APIs to check all the content and activity - including information in applications - pertaining to the query before storing it in a customer's cloud environment. The data stored is then fed to large language models (LLMs), which have been trained on that particular enterprise's data,
As AI accelerates code production, it is widening a bottleneck in the far larger "after-code" phase of software development - the testing, security checks, and deployment work that still consumes nearly 70% of engineering time. Harness's tools help automate this sprawling, error-prone layer, even as enterprises grapple with rising AI code volume and the risks of shipping even a single line of faulty software into production systems.
Context windows of AI models, which indicate the ability of a model to "remember" information, have increased over time. However, researchers have suggested new ways to increase long-term memory of AI models, as they often can't hold context over several sessions. 19-year-old founder Dhravya Shah is attempting to solve problems in this area by building a memory solution, called Supermemory, for AI apps.
While tech and AI giants guard their knowledge graphs behind proprietary walls, a more open model is quietly powering innovative projects from São Paulo to Nairobi. Wikidata, the collaborative backbone behind Wikipedia's structured data, has become the world's largest free knowledge database. Lydia Pintscher, who leads the Wikidata project at Wikimedia Deutschland, oversees this enormous experiment in open collaboration. More than 25,000 contributors across 190 countries have built a database containing 116.6 million data points, edited nearly 500,000 times daily.
The structured inference results into triples enable a clear relationship description between materials by defining core labels, which allows for improved data retrieval and analysis.
The FMKG represents a significant advancement in materials science by utilizing natural language processing to create a knowledge graph that facilitates entity extraction from extensive scientific literature.
