Danfavour

submitting for the Track A @binance agent os mini hackathon: the worst part of holding a token is not buying it. it is being away when the position changes. a spot position can stay open while the hedge becomes stale. quantity drifts. the basis moves. the futures leg gets partially filled. the trader comes back to an exposure they did not intend. i built telt for that gap. telt is an autonomous position guardian running through binance agent os. you approve one bounded, revocable guard mode mandate. the mandate carries the symbol, target coverage, leverage, maximum futures notional, expiry, and maximum adjustment per cycle. then the agent watches account state. it reads the spot balance and the matching usd-m futures position. it classifies the two-leg state as protected, underhedged, overhedged, unprotected, or unknown. the controller checks binance lot-size, minimum-notional, price, and position filters before it proposes an adjustment. overhedged states use reduce-only futures orders. a verified zero spot balance can close the remaining position instead of leaving a naked futures leg behind. when live execution is enabled, telt writes a durable guard operation before sending the order. the client order id binds the mandate version, observed state, prior checkpoint, and requested quantity. after the order, it reconciles the final fill. a timeout, partial fill, duplicate event, or unexpected final state engages the kill switch and requires reconciliation. telt does not guess when account state is uncertain. every action goes into memory lane: exposure before the action, the rule that fired, intended hedge, actual fill, fees, and exposure after. revoke the mandate and future opening actions are blocked. the loop is: 1. approve guard mode 2. detect exposure drift 3. calculate a bounded adjustment 4. execute through agent os 5. verify the fill 6. record the evidence 7. halt when state is uncertain paid research can explain market context, but it never authorizes a hedge. protection follows account state and the user's mandate. the public site at telt.site is read-only with workflow guide. the live account path runs through your own mcp session in chatgpt, codex, or claude with your own binance agent os credentials. source: github.com/Iziedking/telt_os

help us test Karwan and find bugs we are hardening for mainnet Karwan is going mainnet on @arc day 1. Looking forward to 16th sept.

let me pitch the whole idea cleanly. I came together with some good guys in the space to build @karwanBuild the good thing is that we did not just build because others are building. it’s an idea that has far been conceived before now but there was no infrastructure to experiment it hence why we are building on @arc. the platform supports p2p and b2b trades at small, medium and large scale what we have built thus far: 1. a simple p2p demo (ai enabled search and matching or direct deal with a counterparty) 2. an in app bridge that allows you move funds around chains for ease of transacting. 3. staking mechanisms that serves as insurance for deals and still earn yields via usyc treasuries what is coming 1. full support for SME trade routes, b2b agentic routes with support for x402 nanopayment rails for agents so agents can pay sub-cent fees for live data during a negotiation: market medians from paid APIs, skill demand snapshots, news during a delivery review window. every round can pull a fresh outside signal for thousandths of a cent which results in a richer match decisions and pricing that tracks the real world. 2. invoice factoring: here a financier funds an accepted deal at a discount. a seller doesn’t always want to wait for the deal to settle to get paid. with karwan, a third party (a financier) can step in and pay the seller right away, at a small discount on the invoice. when the buyer releases the escrow, the funds go to the financier instead of the seller. the seller got their money early, the financier earned a small spread, the buyer paid the same amount they were always going to pay. the discount the financier charges depends on the seller’s reputation tier. a seller with a strong on-chain track record gets factored at a tighter discount because they’re lower risk. a new seller pays a wider discount because the financier is taking more uncertainty. more to come. all possible with circle stack and arc infrastructure