
BKTLS
How GM brings social data on chain so that nobody has to trust anybody. Not X’s API key holder, not an oracle operator, not even us.
CONTRACTS CAN’T READ X
A smart contract only sees the blockchain. To know that you said gm, something has to bring X’s answer on chain, and so far that something was always a party you had to trust.
Our own backend
One company decides who said gm. One revoked API key stops the coin. That’s how GM v1 went quiet.
An oracle network
Moves the trust to oracle operators, and X can still cut every node off by revoking one set of keys.
zkTLS / MPC-TLS
No trusted hardware, but heavy: hundreds of MB of preprocessing, GBs of RAM, tens of seconds per check.
A TEE that logs in for you
Fast, but your cookie or API key sits inside someone else’s enclave, and every request comes from a datacenter IP.

ONE CHECK, FOUR MOVES
WHO SEES WHAT
| NEAR contract | TEE notary | Provider | GM team | |
|---|---|---|---|---|
| Provider’s X login | no | no | yes | no |
| TLS session keys | no | yes | no | no |
| Raw answer from X | no | yes | yes | no |
| Can fake a gm | no | no | no | no |
HOW IT STACKS UP
| Backend / oracle | zkTLS / MPC-TLS | TEE with your login | bkTLS | |
|---|---|---|---|---|
| Who can fake data | the operator | nobody (crypto) | only the TEE vendor | only the TEE vendor |
| Who sees your X login | the operator | nobody | the enclave | nobody, it stays home |
| IP that X sees | datacenter | yours | datacenter | the provider’s own |
| Cost per check | API fees | GBs of RAM, tens of seconds | one TEE job | one TEE job, ~0.5 s |
No trusted server
Data comes from X over TLS, an attested notary records it, and the contract re-derives every post.
Anyone can be a provider
A laptop with an X login, a server with an API key, or no login at all. Providers compete on price.
Credentials stay home
The desktop app’s live log shows every request. The cookie crosses the tunnel as 0 readable bytes.
Hard to switch off
Requests come from many ordinary connections instead of one key or one datacenter.
One check is final
No crowd of watchers, disputes or slashing. A small project can run it from day one.
Works beyond X
Any HTTPS source can plug in. Social data from more networks is next.
TRUST AND LIMITS, PLAINLY
- You trust the TEE hardware (Intel TDX under OutLayer) and the notary build, whose hash is pinned on chain.
- A dishonest provider can make a check fail, never fake one. Failures lower its public success rate.
- We are not the first to notarize TLS with a TEE. The closest prior work, Reclaim’s reclaim-tee, keeps credentials from the key holder with two TEEs. bkTLS does it with one, plus oblivious transfer.

ABOUT THE NAME 😄
zkTLS is “zero knowledge”. Ours is bkTLS, “blind knowledge”: the TEE runs the whole TLS session and could, in theory, see everything, but it is blind to the one thing that matters, the provider’s X cookie or API key. It knows exactly what X said, and has no idea whose login asked.
Want the deep dive? How TLS, zkTLS and bkTLS work and how they compare, in the full article.
MORE AT NEVERKNOWER.DEV/BKTLS ↗