Roughly 80 percent of GitHub AI projects are abandoned within a year of creation. That single number should reframe how you read every breathless product announcement you see on social media. GitHub star growth is a well-documented vanity metric — it attracts investors and generates buzz, but it tells you almost nothing about whether a tool actually works. The researchers and analysts tracking open-source AI most closely, including those studying fine-tuning trends among startups, have noted that star counts and real adoption are frequently disconnected signals. While community traction tools like Papers With Code are useful, GitHub offers unique insights into project viability through developer activity and repository health metrics. Here is what makes GitHub uniquely valuable, Shubham: it shows you what developers are actually doing, not what companies are claiming. Corporate marketing is optimized for perception. A GitHub repository is a live artifact — commits, pull requests, issue response times, and contributor counts are all visible, and they cannot be faked at scale. The success of open-source projects hinges on active developer engagement and consistent contribution patterns. That means the health metrics that matter are contribution regularity, the number of active committers, response time to issues, and release cycle consistency. A repository with 10,000 stars but a 90-day average issue response time and three total contributors is a warning sign, not a success story. Contrast that with a project showing weekly commits, a growing contributor base, and active communication channels — that is a tool the builder community has genuinely adopted. Interpreting spikes in repository activity, such as forks and new contributors, can signal a project's transition from experimental to practical use. This is where GitHub diverges sharply from press releases. Fine-tuning open-source models is a rising trend among startups seeking competitive edges, and you can see that trend forming in real time by watching which model repositories accumulate fine-tuning forks before any newsletter covers it. Open source also keeps proprietary AI labs honest; when an open model matches a closed one on key benchmarks, the repository activity around it tells you that before the analyst reports do. There are real limitations here, Shubham, and ignoring them will cost you accuracy. Workflows successful in one open-source context often fail in others — a repository thriving in one ecosystem may not translate. Modern software depends on millions of open-source libraries, creating dependency chains that make raw activity metrics misleading; a spike might reflect a dependency update, not genuine adoption. Security vulnerabilities in open-source dependencies are a persistent community challenge, meaning high activity can sometimes signal a crisis rather than momentum. Use GitHub data as a directional signal of developer sentiment and project viability, complementing insights from community traction tools and curated newsletters. The core skill this lecture is building, for you specifically, is learning to read developer sentiment as a separate and often more honest data stream than corporate announcements. Track repository health metrics — committer count, contribution regularity, issue response time, release cadence — not just star totals. Watch for activity spikes in fine-tuning forks and new contributor surges as leading indicators of real adoption. Open source keeps the AI field transparent in ways that no press release ever will. When you learn to read that transparency, Shubham, you stop reacting to hype and start tracking what builders actually trust.