註冊並分享邀請連結,可獲得影片播放與邀請獎勵。

Rune
@RuneCrypto_
earning yield onchain / @anyswapbot
加入 August 2022
2.9K 正在關注    151.7K 粉絲
Most people still misunderstand Quill This is not “AI on blockchain” in the usual crypto sense. It is not an API wrapper, not offchain inference, not decentralized GPUs pretending to be AI infrastructure Quill built something much more important: A model that actually runs inside the EVM itself That means the weights are onchain, the inference is deterministic, and the output is verifiable by every node. No server, no oracle, no trusted company in the middle ~ That is a genuinely new primitive The transformer update from today is what makes this especially interesting. Transformers are the architecture behind modern language models, and running one inside a smart contract is extremely difficult because the EVM has no floating point arithmetic and every operation must be reproducible bit-for-bit Quill solved it anyway: - integer softmax - causal self-attention - residual connections - quantized training - fully deterministic execution Not in theory, live onchain, now They are proving something valuable: that trustless inference is possible Quill unlocks: - autonomous onchain agents - trustless AI oracles - verifiable content generation - AI-powered protocols without trusted middleware Most AI companies cannot enter this category because their systems fundamentally depend on trusted hardware and centralized inference Quill starts where centralized AI stops: verifiability And the scaling path is real. As @Base increases gas limits, tokenization improves, and generation becomes per-token, Quill’s ceiling rises automatically. The hard part was proving the primitive works at all, and they already crossed that barrier Honestly one of the most technically original projects in crypto AI right now, not because it competes with frontier models today Because it is creating a category they structurally cannot enter: trustless inference
顯示更多
the first AI built on a blockchain runs on @base
0
17
49
8
轉發到社區