#nil $NIL @Nillion
been poking around nillion for a few days and i'm still split on it.
the problem is real, which is rare. blockchains are public by design, so you can't put trading strategies, health data or an ai agent's memory on one. nillion's fix is blind computation: the network works on your data without ever seeing it. mostly mpc, so data gets split into shares across nodes and nobody holds the whole thing. more practical than fhe, which is still painfully slow.
but... they added tee-based compute (nilcc) last october, and that's a different trust story. mpc says trust the operators won't collude. tees say trust the chip maker. calling it all "blind computation" blurs that a bit.
then the node count. their own roadmap said ten nodes, six in one enterprise cluster. that's not exactly "permissionless." and the usage numbers, 750 million secrets stored, 1.2 million inference calls, are self-reported. easy to pump. partnerships with coinbase's dev platform, near ai and a meta paper are nice, but logos are cheap in crypto.
token side, i can't tell if the network needs NIL or NIL needs the network. last number i saw was around three cents and a ~$15m market cap, but that was august so it's probably moved.
and everyone in decentralized storage and compute asks the same thing: why would a company leave aws? privacy is a better answer than price, but the fhe crowd and tee networks are chasing the same customers.
interested, not convinced. the one thing i'd watch is whether a boring non-crypto company quietly uses it. usage doesn't lie.
not financial advice, just a guy typing.
been poking around nillion for a few days and i'm still split on it.
the problem is real, which is rare. blockchains are public by design, so you can't put trading strategies, health data or an ai agent's memory on one. nillion's fix is blind computation: the network works on your data without ever seeing it. mostly mpc, so data gets split into shares across nodes and nobody holds the whole thing. more practical than fhe, which is still painfully slow.
but... they added tee-based compute (nilcc) last october, and that's a different trust story. mpc says trust the operators won't collude. tees say trust the chip maker. calling it all "blind computation" blurs that a bit.
then the node count. their own roadmap said ten nodes, six in one enterprise cluster. that's not exactly "permissionless." and the usage numbers, 750 million secrets stored, 1.2 million inference calls, are self-reported. easy to pump. partnerships with coinbase's dev platform, near ai and a meta paper are nice, but logos are cheap in crypto.
token side, i can't tell if the network needs NIL or NIL needs the network. last number i saw was around three cents and a ~$15m market cap, but that was august so it's probably moved.
and everyone in decentralized storage and compute asks the same thing: why would a company leave aws? privacy is a better answer than price, but the fhe crowd and tee networks are chasing the same customers.
interested, not convinced. the one thing i'd watch is whether a boring non-crypto company quietly uses it. usage doesn't lie.
not financial advice, just a guy typing.