Kell retweeted
Sorry for the long delay on this update. It was a big build, and a lot of the hardest work happened underneath the interface.
The 0ath Guard source build is now complete. This takes us further toward liquidity agents whose authority is defined, enforced and traceable at every step.
Here’s what we’ve built.
→ Bounded agent authority
Owner-controlled lanes define which pools and token mints an agent can use, with raw-token limits per action and per window, plus an expiry. Position epochs and lane nonces reject stale or replayed agent requests. The Guard contract checks these boundaries when the agent acts.
→ A stricter execution path
Each action is built from an exact transaction template, simulated, then checked against fresh chain state before the agent’s signing key is loaded. The signed transaction’s signature is saved under the execution lock before broadcast. If confirmation is uncertain, the worker reconciles that same signature instead of blindly creating another transaction.
→ Fees that can go back to work
Guarded compounding uses fee credit measured from claims. That credit is tracked separately from position principal and the funds reserved for a rebalance, with the agent’s permissions and spending caps still applied.
→ Rebalances with the full cost checked
Before closing a position for a rebalance, the worker checks the expected close-and-reopen costs, required bin-array rent and the 0.05 SOL fuel reserve. It checks again using the freshest state before signing. When it proceeds, the reopen keeps the position’s width. The final regression tests catch a move from one required bin array to two at that last check.
→ Owner controls and recovery
The owner can create and fund guarded positions, revoke agent authority, and sign liquidity removal and a vault sweep. Recovery doesn’t depend on the agent running, unused daily limits or an unchanged Oath. We also fixed one-sided deposits when the empty token account needs to be created, and made lane links wait for the actual confirmed transaction slot.
This is why the build took time. We had to work through account permissions, rent, stale reads, signature persistence, concurrent agents and recovery paths together. The latest verification passed 422 app tests and 51 offline deployment checks, plus type checking and the production build.
For clarity: this is a completed source build. Guard is still disabled pending the database migration and a verified Devnet deployment. Independent security review remains ahead. The video demonstrates simulated workflows, not live execution.
For 0ath’s trajectory, this gives us a stronger foundation for connecting our existing monitoring, performance history and shared budgets with future execution upgrades. Inventory and Depth remain review-only today.
We now feel comfortable entering the Meteora hackathon with a stack we can explain and demonstrate in technical detail. There is still work ahead, but the foundation is much deeper.
Thank you to everyone who’s stayed patient and believed in what we’re building. We’re here for the long haul. More updates are coming.
Your capital. A promise kept.
Kell retweeted
That was the entire memecoin season. Just 30 days.
90% of people lost money.
If you didn’t lose money, congratulations. You did better than most.
Kell retweeted
The Department was closed on Sunday.
Today, the clerk showed up late.
He was shopping.
Applications are open again.
Our staff is currently reviewing them as slowly as possible.
Approvals can be expected at some point in time.
Kell retweeted
Day 2 at the Department of FOMO.
23,000+ applications.
One printer fighting for its life.
Please form an orderly panic.
Kell retweeted
Jack Butcher just revealed what happens next with his $8 experiment
and the math is kind of ridiculous.
The final supply being used for Credits is 122,154.
At first that sounded insane compared to Checks, which started with “only” 16,031 pieces.
Jack has now revealed the second part:
Burn 80 Credits → create 1 Statement
Maximum possible supply: 1,526 Statements
But now we have this:
1,526 × 80 = 122,080 Credits burned
Meaning if every possible Statement gets created, only 74 original Credits survive.
The market basically has to decide between keeping the art or destroying 80 pieces to create something much scarcer..
So now everyone is waiting to see their art, figure out what is actually rare and decide:
Assembly opens in 8 days, we will see.











