#nil $NIL @Nillion
I went looking for what 643.8M "documents" in nilDB actually amounts to, and the storage figure stopped me: 297.86 GB total. That's roughly 460 bytes per document, closer to a database record than the high-value AI-agent data the pitch describes.
The same dashboard shows 112.7K users, 1.4M inference calls and 2.6B AI tokens processed. That's about 1,850 tokens per call, which reads like real usage rather than empty pings, though aggregates can't prove it.
The staking number is where I'm least sure. 20.6M NIL staked across 37 Blacklight nodes. At the 70,000 NIL minimum, node collateral accounts for only about 2.6M of that. Either most of the stake sits elsewhere or I'm reading the metric wrong.
Two caveats. These usage figures are self-reported, and I couldn't find an independent source for them. And while the credits model burns NIL for nilDB and nilCC, nilAI is paid through Stripe, with no published burn or revenue total. Meanwhile the L2 contracts for on-chain staking and direct access are still slated for later in 2026, and about 48M NIL unlocks between October and December.
So the network is probably being used. Whether that usage pulls on the token is the open part.
If nilAI revenue settles off-chain, what would you accept as evidence that NIL demand actually follows usage?
I went looking for what 643.8M "documents" in nilDB actually amounts to, and the storage figure stopped me: 297.86 GB total. That's roughly 460 bytes per document, closer to a database record than the high-value AI-agent data the pitch describes.
The same dashboard shows 112.7K users, 1.4M inference calls and 2.6B AI tokens processed. That's about 1,850 tokens per call, which reads like real usage rather than empty pings, though aggregates can't prove it.
The staking number is where I'm least sure. 20.6M NIL staked across 37 Blacklight nodes. At the 70,000 NIL minimum, node collateral accounts for only about 2.6M of that. Either most of the stake sits elsewhere or I'm reading the metric wrong.
Two caveats. These usage figures are self-reported, and I couldn't find an independent source for them. And while the credits model burns NIL for nilDB and nilCC, nilAI is paid through Stripe, with no published burn or revenue total. Meanwhile the L2 contracts for on-chain staking and direct access are still slated for later in 2026, and about 48M NIL unlocks between October and December.
So the network is probably being used. Whether that usage pulls on the token is the open part.
If nilAI revenue settles off-chain, what would you accept as evidence that NIL demand actually follows usage?