BREAKING: 🇺🇸 President Trump is set to make a “huge” announcement today at 5:00 PM ET.
Sources are speculating it could involve plans to reopen the Strait of Hormuz and a possible new peace deal with Iran. Markets could see major volatility if confirmed.
@OpenGradient has been sitting in my tabs for a few days and the thing that keeps surfacing is the asynchronous proof model.
The design decision that gets less attention: inference runs first, settlement happens after. Full nodes don't re-run the AI computation in real time they verify proofs separately after inference is already complete, which lets the network keep speed comparable to traditional cloud services
While still settling a verifiable record on-chain. (NFT Evening) That sounds like a footnote in the docs but it's actually the whole architecture. Real-time trading agents and DeFi risk models can't hold open a connection while a zkML proof computes. So HACA just decouples them.
This week, OPG dropped 19.4% over seven days and touched within 1% of its all-time low near $0.1278, while 24h volume climbed 18% from the day prior to roughly $32.8M. That divergence is worth sitting with. Volume rising near a floor doesn't confirm accumulation could be liquidations clearing but it does confirm the market isn't just walking away quietly.
Meanwhile the network side doesn't show signs of slowing: 2M+ verifiable inferences logged, with 500K+ zkML proofs and TEE attestations settled on-chain. Protocol activity and price discovery seem to be operating on completely separate timelines right now.
I'm still not sure whether the async proof design matters at scale or only makes sense for narrow latency-sensitive use cases. How much of that inference count is real production load versus teams testing their own pipelines?
$SUI is trying to bounce from trendline support. This looks like a clean dip-buy setup only if it keeps holding above $0.678 – $0.680. Current SUI is around $0.687.
Been sitting with @OpenGradient for a few days, and the part that keeps pulling my attention isn't the AI narrative it's the architecture decision behind OPG core design.
OpenGradient separates execution from verification entirely. Inference nodes run the model and return a result fast. Full nodes verify the proof separately, asynchronously. Those two things don't wait on each other.
What made that concrete for me was checking on-chain behavior this week. Even as $OPG dropped roughly 19% over the last seven days, 24-hour volume ticked up around 18% on June 26. Price down, interaction up.
That's a minor divergence, but on Base where inference payments settle in real time through Permit2, it suggests some portion of that volume isn't purely speculative requests are actually hitting the network.
The "verify later" design is what unlocksthis. If every inference had to wait for a zkML proof or TEE attestation before returning a result, latency would kill any real use case. By decoupling the two phases, OpenGradient lets the user get an answer now and the network prove it was honest afterward.
Not a new idea OpenGradient offers a verification spectrum, including zkML proofs and TEE attestations, allowing developers to choose their own balance of cost, speed, and security but applying that tradeoff framework to AI inference specifically is where it gets interesting.
I'm still not sure how proof backlog behaves under real load. Does verification keep pace with execution at scale, or does that gap widen quietly? That's the part I'd want to stress-test before drawing any conclusions.
BREAKING: The ceasefire is already under pressure.
The 🇺🇸 U.S. has struck Iranian missile sites, drone depots, and radar systems after Iran reportedly attacked a ship exiting the Strait of Hormuz marking the first U.S. strike on Iran since the MOU was signed.