Imagine telling someone in 2024 that not over-relying on synthetic data is a good thing.
The reality is LLM outputs are biased towards the high-density regions of the training distribution.
Synthetic data is a lossy reconstruction of reality. That's great for augmentation and distillation. But if models increasingly train on the outputs of other models, we're effectively recompressing a compression. The first thing that disappears is the long tail (rare observations and genuinely novel information), and much of what we call intelligence is the ability to reason about those exact things.
The next GRASS Token Holder and Network Participant Call will take place on July 7. The call will provide an overview of Grass network progress, the development roadmap, business milestones, growth highlights, and more.
To promote equal access to information, the call will be open to the public. Questions may be submitted in advance for consideration during the session.
Further details, including the exact time and registration information, will be shared soon.
The network is more reliable than it's ever been as our success rates and core performance metrics have increased across the board.
This is a result of improvements to our detection and filtering capabilities that remove bots, bad actors, and low quality nodes from the network. Our efforts to prevent bad actors from gaming the system have led to stronger and more sustainable network performance.
Our goal has always been simple: reward real users who support the network with bandwidth that the network can actually use.
Real rewards for real users.
Nom: "I can be in Miami for 24 hours."
The Dog: say less.
2 podcasts, fireside chat on @consensus2026 stage, BONK Community Event and then a red eye home 👇