Daily Briefing — April 21, 2026

A fake GitHub star costs six cents. A seed round unlocks one to ten million dollars. A peer reviewed study out of Carnegie Mellon, NC State, and Socket analyzed 20 terabytes of GitHub metadata and found roughly 6 million fake stars spread across more than 18,000 repositories, purchased by around 301,000 accounts. By mid 2024, one in six repositories with 50 or more stars showed signs of manipulation. The scale points to a mature shadow economy running in broad daylight.

The mechanism is simple. Star marketplaces sell engagement at scale, repositories climb the trending rankings, developers and investors see social proof, and funding or adoption follows.

If you use open source tools at work, evaluate libraries for your stack, or make hiring decisions based on someone's GitHub profile, you have been operating on polluted signal. Star counts, contributor graphs, and trending lists are not neutral metrics. They are marketplaces. The developers and teams who understand this will make better decisions. The ones who do not will keep getting played.

The Guardian ran a podcast following a new teacher navigating a classroom where the students already use AI fluently and the institution has no idea what to do about it. The teacher describes it as "downing a coffee in the middle of a panic attack," which is maybe the most honest summary of what happens when a legacy system meets a tool that does not care about your curriculum plan.

So what happens to the humans whose job is to teach when the thing they are teaching can now be generated on demand. The teacher is being repositioned, and nobody gave them the memo about what the new job actually is.

Here is a slow moving but genuinely fascinating collision happening inside one of the most change resistant professions on the planet. Lawyers are famously bad at tech. We have the Zoom cat incident on record as proof. But something bigger than embarrassment is now at stake. A quirk in how malpractice liability works means lawyers could soon be legally obligated to use AI tools, not because they want to, but because failing to use them might constitute professional negligence. The argument goes like this: if AI can reasonably be expected to catch an error, find a precedent, or flag a risk that a human missed, then not using it becomes the problem.

The same profession that got fined and sanctioned for trusting AI too much may now face consequences for trusting it too little. What a curve. Every profession that absorbs a new tool goes through a phase where misuse gets punished before the norms settle. Legal is just going through it louder and later than most.

The question is shifting from whether you use these tools to whether you use them well enough.