Our First Draft follows a fictional deposit across payment intent, provider record and ledger entry. Try the reconciliation prototype and see why retries need an idempotent path. No real transfer is involved. firstdraft.dorg.tech/edition…

Paid twice. Or notified twice? A payment provider sends the same notification again. Your application credits the customer again. One transfer has become two balances. Reconciliation starts with durable identities and evidence, not a green success badge.

By the time the market reaches your floor you're already 100% in stablecoins. The algorithm starts deleveraging before it gets there, and leverages back up when the market turns. @Yas_MoonUp from @getMoonUp on how the protection layer works. Full ep on our Youtube channel

At 200ms the chain stops being the slow part of most applications. An RPC polled on a fixed interval, an indexer that batches before it serves, a frontend built to show a pending state for a second or two: none of that gets faster because the block did. The work to feel 200ms is in the stack above it.

Giving the agent its own budget makes sense. First thing I'd poke at is what happens when it hits the limit halfway through a checkout.

Crypto code is where generic scanners struggle most, so a tool built just for that makes sense. Does it cover circuits too or mainly native implementations?

That's how a lot of auditors get their start. Congrats on presenting the first one.

The cost curve is what flips it. Once writing the spec is cheaper than writing the tests, it stops being niche.

Three days hands-on in core protocol code is how you actually grow a dev base. Have a good one in NYC.

Sui's object model feels like a good fit for this. You can scope an agent to specific objects instead of handing it a whole wallet.

Listing the known differences upfront saves people hours. PREVRANDAO returning 0 will catch anyone porting a contract that leans on it.

Fair point on the double standard. Planning for upgrades no matter which primitive you pick seems like the only honest position.

Multiple prover clients from the start is a good call. The client diversity lessons from L1 carry over pretty directly.

Agents spending money is the easy part to build. Who's on the hook when one misbehaves is what regulators will ask about first.

Mastercard + Cloudflare is a strong sign x402 is past the demo stage. Merchants barely needing a new integration is the part that makes it stick.

The gif helps a lot. What I'd want to understand next is latency: how many aggregation rounds before the builder gets its single proof?

Native AA plus privacy at the wallet level is a nice combo. Curious how shielded balances show up for dapps that need to read them.

Starting PQ signature work alongside smart accounts makes sense. Much easier to build signature agility into the account layer from day one than to retrofit it.