This article outlines workflow principles for building software with AI agents rather than traditional sprints. Key strategies include running multiple agents continuously, auditing apps with customer personas, and converting successful experiments into reusable skills.
This six-step growth playbook outlines tools used from analytics to product communication, including Mixpanel, FullStory, UserTesting, VWO, Pendo, and HubSpot.
We need ways to prove that code is free from bugs. Formal software verification, which could provide this guarantee, has always required years of expert effort, but LLMs can now produce formal proofs at large scale. This enables new software development workflows that include proven security guarantees. Formal verification will likely become an essential part of secure software development as model capabilities continue to improve.
In enterprise product development, choosing between buying tools and building them internally has become a user experience decision driven by AI. While coding agents reduce initial build costs, ongoing maintenance, data silos, model retirement, and user context switches introduce hidden complexities that product teams must manage.
Harvey's marketing team used agents to make a tailored product film for each of hundreds of target accounts.
Anthropic employs around 3,500 staff, but its coding output is closer to that of a company with 35,000 to 70,000 staff. Big tech companies like Google and Meta moved off Git a long time ago because it did not scale for them. Both built their own custom solutions, just like Anthropic did with Scion. This post provides some details about how Anthropic's custom source control system works.