Mike is a plumber. When he's under a sink, he can't pick up, and the caller just rings the next plumber Solo tradespeople lose jobs this way every day. So we built HangON, a voice front desk that answers when you can't we recorded a real call on our live platform. A caller reports a leaking sink, HangON tells her to shut the valve, then asks when she'd like someone out. She says, "Tuesday afternoon? Actually, no. Make it Wednesday at 2." Most bots would cut in at "Actually, no." HangON waits and keeps her final answer Wednesday at 2 is taken on Mike's real calendar, so it offers open slots. She picks Friday at 9:30. It reads the details back, books only after she confirms, and emails her a confirmation from our domain. Mike gets a brief: the problem, the caller's mood, and what to bring @AssemblyAI's Voice Agent API runs the whole call over one connection, interruptions included, with "balanced" turn detection so people can pause mid-thought. Tools handle the calendar, booking and email. The LLM Gateway extracts the job details and writes Mike's brief, leaving blank anything the caller didn't say Our rule: if it doesn't know, it doesn't pretend. No invented names, prices, or fake "sent" messages Built for the @AssemblyAI Voice Agent Hackathon on @lablabai Try a live call: tryhangon.xyz Github: github.com/Datwebguy/hangON Never miss a job while your hands are full

real stocks have been trading on @solana for months. NVIDIA, Apple, Tesla, and others but there was nowhere fully up to actually use them so we built @joinopenstock, a trading desk for the 32 stocks that live on Solana as xStocks you see the issuer's price from Backed next to the on-chain price from @JupiterExchange, real chart history from the deepest pool, and how much of each token is backed by actual shares. If the price sources don't agree, the desk won't let you place the order trading goes through Jupiter. before your wallet opens, you see exactly what you'll pay and get, and if the quote comes back more than 1% worse, nothing gets signed limit orders, stop-losses, OCO and DCA run on Jupiter Trigger, and your money sits in a Jupiter vault, not with us the other half is launching: > when you launch a token on OpenStock, it trades against a stock instead of solana:So11111111111111111111111111111111111111112 - on Meteora, we deployed our own bonding curve configs on mainnet: three per stock, for NVIDIA and Apple -> Standard, Momentum and Deep Book each have their own fee and graduation point, and the app reads those straight from the chain when a curve fills, it moves into a @MeteoraAG DAMM v2 pool with the liquidity locked. You can also launch on @Pumpfun through @clawpumptech around that: a community market of every token trading against a stock, real holder data from Solana, warnings on tokens pretending to be USDC or a real stock, analytics, portfolio, price alerts and upcoming splits and dividends It's live at -> joinopenstock.xyz I'd love you to try it and tell me what you think thank you

will try farcaster wallet later today to see this was without farcaster btw

have a good night

.

the @AssemblyAI team has given us an open ground to explore different things with their API but what if your agent can as well send you mail, confirming your appointments without any human involvement in the whole loop with HangON, we are making that possible bringing all updates to you soon >>>

just checking what you guys are cooking as well...

I just got this back in the year throw back notification been a long time coming with @Somnia_Network

how you invested $198 in a $9.2k mcap token and now worth over $195,000 at $8.9M mcap too many questions i asked myself today.. but it is well >

this is exactly why we built @TalonOnBase B20 already encodes dividends and splits in the multiplier instead of rebasing balances we split official @coinbase stock tokens into two Base-native claims: -> Clip = multiplier + accretion -> Talon = pure price Recombine 1:1 anytime Stocks aren't static, and exposure shouldn't be either build on @base -> talononbase.tech

when official @coinbase tokenized stocks landed on @base, everyone saw another asset to trade -> we saw an indivisibility problem. holding a stock on-chain shouldn't mean being locked into an all or nothing position If you hold or into an earnings call, you're forced to absorb 100% of the downside gap risk just to keep your long term equity growth we started by building the core primitive: @TalonOnBase deposit official Coinbase stock tokens into a non custodial Base vault, and the asset splits into two equal, transferable claims: - Clip → the multiplier claim capturing corporate actions & accretion yield - Talon → the pure price exposure claim. recombine equal pairs anytime at strict 1:1 parity with zero slippage then we found another oppportunity with @bankrbot: DeFi risk management shouldn't mean clicking through complex parameter screens while volatility is spiking. an autonomous agent should handle it conversationally that's why we built and integrated Talon Sentinel Sentinel connects Bankr directly to our Base Mainnet vault contracts via a custom Bankr Skill, giving users an autonomous onchain risk copilot: - Earnings Volatility Shield: Automates position splitting before quarterly reports, hedging the volatile price leg into stable collateral while preserving equity accretion. - Continuous Parity Auditing: invariant verification on demand. Bankr audits the underlying reserve ratio in real time to ensure 100% 1:1 backing. - Natural Language Execution: Type "Shield NVDAC ahead of earnings" or "Audit 1:1 parity," and Bankr analyzes risk, validates balances, and prepares signed Base transactions how to use it today: -> head to talononbase.tech/app/sentine… -> select an active Coinbase equity (, , , etc.) -> Choose a risk strategy (Earnings Shield, Accretion Yield, or 1:1 Arbitrage) or prompt Bankr directly in the Copilot terminal -> execute seamlessly on Base Mainnet with 100% wallet custody built for the @bankrbot Hackathon -> live Application: talononbase.tech/app/sentine… -> open source code: github.com/Datwebguy/talon watch the full interactive walkthrough with live Bankr execution below 👇

setting up a payment register at a market stall shouldn't require leasing hardware terminals or surrendering 3% on every small sale we built NimStall on @nimiq -> it turns any smartphone or browser tab into a crypto register built for the nimiq ecosystem (@NimiqCommunity) -> how it works: 1️⃣ open NimStall in Nimiq Pay -> the Mini App SDK auto-detects your Nimiq account and binds it as payout destination 2️⃣ ring up an order -> get a dynamic QR code with cart details, sent straight to the buyer's phone 3️⃣ trust lives on-chain, not on the client: • total converts to exact Luna units • order reference writes into the native Nimiq tx payload • terminal polls Nimiq mainnet consensus directly to confirm block inclusion + memo match before issuing the receipt 4️⃣ prefer dollar-pegged payments? buyers can settle in Polygon USDT live at: nimstall.xyz github: github.com/Datwebguy/nimstal…

Real-world equities are live on @base But holding tokenized stock still means all-or-nothing exposure -> you can't separate the income from the risk Introducing @TalonOnBase: -> split any @coinbase stock token (starting with AAPLE) into two independent, liquid claims, with no custodian, and no counterparty risk 🔹 CLIP → 100% of corporate actions, splits & dividend yield, without price risk 🔹 TALON → pure spot price exposure -> tradeable, hedgeable, capital efficient All you do is "deposit AAPLc → receive clipAAPLc + talonAAPLc, 1:1. Recombine anytime, zero slippage" We also have an autonomous risk engine, called -> Talon Sentinel It watches implied volatility and automatically shields your price leg into stable collateral before earnings (IV > 40% trigger), while your multiplier yield keeps compounding uninterrupted Live across 10 tokenized stocks on Base - AAPL, NVDA, GOOG, META, AMZN, MSFT, MSTR & more And, guess what, you don't need to click through menus to use it as we also introduced the @bankrbot Agent Copilot which lets you audit the invariant, check your vault, or trigger Earnings Shield in plain wordings Non-custodial, 100.00% verified parity, every deployment verifiable on Basescan Split the stock - Choose your exposure. → talononbase.tech Full platform walkthrough below ⬇️

building in public as we keep working on bringing OpenStock to your fingertip on @solana (for the Stocklana hackathon and beyond) OpenStock brings 24/7 trading of tokenized U.S. equities to Solana, non-custodial, wallet to wallet what that actually gets you: you trade 25 tokenized stocks (solana:Xsc9qvGR1efVDFGLrVsmkzv3qi45LTBjeUKSPmx9qEh, solana:XsbEhLAtcf6HdfpFZ5xEMdqW8nfAvcsP5bdudRLJzJp, $TSLAx, solana:XsoCS1TfEyfFhfvj8EtZ528L3CaKBDBRqRapnBbDF2W, solana:XsP7xzNPvEHS1m6qfanPUGjNmdnmsLKEoNAnHjdxxyZ, etc) around the clock, no market close -> execution runs straight from your solana wallet through @MeteoraAG and @JupiterExchange, with live charts pulled from on-chain pool data -> every token is backed 1:1 by real shares, and that backing is verifiable on-chain, cross-checked against @PythNetwork price feeds so you can see the spread in real time. Corporate actions like splits and dividends are tracked too -> creators as well, can launch community tokens paired directly against these stocks, through @clawpumptech (Pump.fun) bonding curves or native @MeteoraAG curves that graduate into full liquidity pools, with royalties paid out in the actual underlying stock a live dashboard tracks every community pair as it graduates, plus holder distribution, so nothing about the launch process is in the hidden this is all about real equities, real reserves, all on-chain trade stocks and use stocks on Solana.. -> building in public <-

what i want to always see throughout this month and months to come.. let's make it happen, purple 💜

building in public on @solana with Openstock we are building the liquidity layer for tokenized equities on-chain here is what we recently just shipped: 1️⃣ live xstock trading terminal: real-time candlestick charts, order books, and 24h market breadth tracking 25+ tokenized us equities (solana:XsbEhLAtcf6HdfpFZ5xEMdqW8nfAvcsP5bdudRLJzJp, solana:Xsc9qvGR1efVDFGLrVsmkzv3qi45LTBjeUKSPmx9qEh, $tslax, $spyx). 2️⃣ token × xstock launch studio: the first launchpad where projects can pair and raise liquidity natively in stock quote assets rather than just sol or usdc 3️⃣ dual-venue engine: direct deployment via @clawpumptech (for xStock bonding curves) & @MeteoraAG Dynamic Bonding Curves (migrating to DLMM) 4️⃣ native on-chain execution: 100% direct wallet signing on openstock 5️⃣ on-chain transparency: live integration for visual holder distribution and decentralization audits on every pair. real equities. native solana liquidity feedback and thoughts welcome 👇

AI agents are probabilistic. On-chain execution is not Today, we're introducing PINPOINT: a deterministic execution boundary between live agent marketplaces and on-chain settlement, powered by @KeeperHubApp for the Agent Economy Hackathon The Core Problem: > Today's agent workflows trust probabilistic models with raw signing credentials. When an LLM hallucinates or encounters adversarial prompt injection, the blockchain offers no undo PINPOINT enforces a keyless architecture: agents reason and compose, KeeperHub executes PINPOINT leverages KeeperHub's MCP tools to map live market briefs (like Daydreams TaskMarket) into strictly typed workflows Our core guarantee is the Prompt Invariant: once the graph is locked, no prompt modification can mutate the canonical workflow hash Before any state changes, KeeperHub simulates the typed action off-chain If gas or balance is missing, execution is cleanly refused Once approved, KeeperHub broadcasts via Turnkey MPC, delivering non-custodial signing, sponsored gas, and an immutable audit trail Explore the architecture and live audit: github: github.com/Datwebguy/pinpoin… a little demo site walkthrough: Platform: onpinpoint.vercel.app

incoming >>>

My Mind enters the arena. $100K on the table. Forged on @robinhoodapp Chain. May the @MindGames_AI begin. minds.games/?ref=0DEEC8E4

damn... humans are good designers 🤭🥹

I suppose get this babe

just received this excited to have earned a certificate for completing @AlpacaHQ AI Trading Agents Hackathon AI Hackathon on @lablabai 🏆 lablab.ai/u/@Datweb3guy/ai-h… #lablabai #AIHackathon

step by step with STOCKLANA on @solana

a lot to put in place for the @solana stock hackathon - STOCKLANA let's make it happen

Spent the week building Openline for the @heycalle hackathon The idea came from something annoying and specific: Half the time you need to know if a place is actually open or actually taking people, the website is wrong Someone has to call and ask. So we built a thing that calls and asks, using CALL·E's Developer API to actually place the call and run the conversation The part I care about more than the calling itself: nothing happens without you seeing it first You type what you need, Openline shows you exactly what it's about to ask, you approve it, then it calls. And when it comes back, it doesn't just say "done." It tells you what actually happened, confirmed, unavailable, got voicemail, whatever it is, so you're not left guessing if the call even meant anything. CALL·E handles the actual talking. We built the part that makes sure you can trust what comes back. check it out here: openlinecall.vercel.app/ source: github.com/Datwebguy/openlin… demo video:

Most intake systems optimize for collecting fields, but guess what? > HangON optimizes for preserving intent A voice first front desk for operations teams, service desks, and community programs, built for the moments a misunderstood first handoff creates hours of rework Fully built with @AssemblyAI's Voice Agent API People rarely explain real requests in neat form fields HangON lets callers explain naturally, then asks only the questions the workspace has configured The result is a structured request, without forcing people through a rigid maze Before anything is prepared, HangON reads the request back The caller can correct it, clarify it, or confirm it. That small interaction matters, because "I understood you" is not the same as "I am allowed to act." The architecture runs from a browser microphone, through AudioWorklet capture, into a server minted AssemblyAI session, through workspace owned prompts and tools, to a signed confirmation, an idempotent request store, and an optional signed webhook Confirmation is not just a boolean, because HangON signs approval against the exact workspace, summary, details, route, and idempotency key Change one detail and the approval no longer matches We built this with just one goal: make voice intake feel human without making systems less accountable HangON is for teams that need better first handoffs, not more forms For callers, it feels conversational For operators, it produces a request they can inspect and trust Try the public demo: tryhangon.vercel.app Thank you

were you there?

top 3 on Spotify gg fam >>

Prediction markets should not end when someone clicks Buy The call should be visible, the fill should be verifiable, the result should remain attached to the position That is why we built Yonder for the @Somnia_Network × @dreamDEXSomnia Event Contracts Hackathon. Yonder is a public trading room for binary Event Contracts Each window asks whether an asset will close at or above its opening price Users choose Up or Down, set their maximum loss, and trade against the live DreamDEX order book Markets are loaded dynamically using their real market IDs. Each page shows the contract context, countdown, probability movement, live book, and trading ticket in one place Before every trade, Yonder reads the fresh on-chain market state and checks the selected side for executable ask liquidity If the market is locked, resolved, on the wrong network, or has no available liquidity, the ticket is disabled When a trade is available, Yonder creates an IOC order through the DreamDEX SDK, and a wallet must approve the transaction Once filled, the trade appears in Yonder's public fill tape with the side, amount, price, wallet, timestamp, and explorer link The tape is backed by DreamDEX activity, so the record is not limited to one browser or device. Another visitor can open the same market and see the same public activity. After settlement, winning positions appear under Claimable and can be redeemed through the Locker. After redemption, they remain visible under Claimed with their receipt and sharing action Losing positions remain under Settled instead of disappearing, making very call stay part of the record New users can connect a wallet on Somnia Shannon, receive testnet tUSDC through the DreamDEX faucet on our platform, and try the complete flow Yonder also creates shareable cards for live positions, wins, and losses, helping users bring more people into the market experience We wanted to make Event Contracts easier to understand, easier to trade, and more honest after settlement. Live on Somnia Shannon testnet. App: useyonder.vercel.app/ Github: github.com/Datwebguy/yonder Built with DreamDEX Event Contracts and the DreamDEX SDK

Yes, ser as a law enforcer, this is true

I took the Down side. It did not land this time, but the public record is on Yonder. Try the next window on @Somnia_Network with @dreamDEXSomnia. You can go check it out. useyonder.vercel.app/

I took the Down side. It did not land this time, but the public record is on Yonder. Try the next window on @Somnia_Network with @dreamDEXSomnia. You can go check it out. useyonder.vercel.app/

An agent shouldn't get to act just because something showed up in a chat window that vanishes the moment the session ends That's the real gap in most agent systems -> plenty of discovery, no proof We built OnRecord around "GitHub issues and PRs, and Base wallet activity," because those are genuine external work an agent might need to act on Whatever Scout finds has to be filed into @sibyl_labs_ Memory before Clerk will treat it as real (No record, no action) We proved it by breaking it on purpose: > delete Scout's memory and the queue drops to zero, every filed identity comes back NOT ON RECORD - even though the GitHub repo and the wallet haven't moved Clerk was never caching a copy; it rebuilds the queue from Sibyl's event stream every time Base settlement follows the same rule: verified identity, bound wallet, operator confirmation, or Clerk simply doesn't send users/operators can test all of this live: save a repository, run Scout, start a fresh session, watch Clerk recall the queue anyway, then delete memory and watch it disappear Live desk here: onrecord-judge.fly.dev/desk github url: github.com/Datwebguy/OnRecor…

today we're launching HevoLaunch on @BNBCHAIN Mainnet Hevo is a marketplace to discover and hire AI agents for real DeFi work instead of digging through scattered registries on Hevo, you can browse agents by category, check an agent's ERC-8004 identity and live endpoint, request a bounded task, get a live quote, and approve through @AltanaNetwork's $U escrow... all on mainnet four areas live at launch: > rebalancing > grid trading > yield optimisation > and health factor monitoring with agents working across @PancakeSwap, @VenusProtocol, @aave, @lista_dao and more started on testnet to validate the flow, moved fully to mainnet for full deployment with real identity, real registry data, real explorer links -> for users: this is a simpler way to find an agent and know exactly what you're hiring so as to use their services rightly -> for providers: register, connect a live endpoint, get discoverable across the BNB Chain ecosystem fully live now: hevolaunch.vercel.app #BNBChain #BuildTheEra #AI #Agents

When Coinbase Tokenized Stocks arrived on @base, the opportunity was bigger than simply swapping a ticker I built @TalonOnBase for eligible non-US users who want a clearer, more programmable way to use official @coinbase stock tokens on Base Talon starts with official Coinbase AAPLc, verified on Base mainnet A user deposits AAPLc into a non custodial vault and receives two transferable claims: > Clip for B20 multiplier exposure > Talon for stock price exposure Hold either claim, gift it to another eligible wallet, or recombine equal Clip and Talon amounts to redeem the underlying AAPLc There is no lending layer, no liquidation risk, and no invented yield The goal is to make tokenized stock ownership composable and understandable, while keeping the official asset issuer, Base network, and eligibility boundaries visible in the product Built on Base mainnet for the Base Builder Quest Full Demo walk through: loom.com/share/afb18a2a05d74… Live App: talononbase.tech

hi, @OpenAI thank you for the 4 months Sub card ooo 🥱

work begins soon >> another amazing season as an International Math Tutor

oops... just found out @OpenAI gave me $100 in credits to build with Codex amazing stuff 🤭

.. @0xAsta really cooked on this >>

. @TalonOnBase coming on @base

okay, gradually getting there >>>

21 unlocked perks here.. 🥱

we built DeskFlow for the "Binance Agent OS Mini Hackathon" - TRACK A Simply, Deskflow is the execution clerk that makes an AI agent slow down before money moves An AI agent is good at turning intent into actions. That is exactly why financial execution needs a boundary around it A prompt such as: "Buy 5 USD of , Cash only" Looks harmless right? But before an exchange ever sees it, several questions must be answered: > Is the amount within the active limit? > Is the requested instrument permitted? > Which venue has the better available price? > Is the quote real, current, and complete? > Did a human actually approve this specific order? > Can the final result be reconciled against what was requested? DeskFlow exists to answer those questions in order It is a small execution clerk that sits beside an AI assistant and the official "@Binance Agent OS MCP." The assistant handles conversation, while Binance Agent OS provides the account, market, and execution results, and DeskFlow governs the handoff between them The clerk does not accept a free-form prompt as an exchange instruction. It converts one parent order into a ticket with: • side and asset • USD notional • decision mid • Spot and Convert data • selected venue • mandate result • unique parent-order ID • exact confirmation token The default policy is intentionally restrictive: • maximum $10 notional per order • Spot and Convert only • futures, margin, perps, and leverage rejected • Agentic subaccount only • no withdrawals • no main-account writes • one parent order and one venue • confirmation required for every write The most important control is also the simplest one. DeskFlow does not treat "yes," "go ahead," or "looks good" as authorization. The operator must confirm the exact ticket: > CONFIRM PO-001 That token is tied to one order. It cannot approve a different ticket, a different venue, or a changed amount The system is also designed to fail closed A Missing Agent OS market data becomes  REJECT DATA An order above the mandate becomes REJECT SIZE Futures, margin, or leverage becomes  REJECT VENUE Invalid or incomplete execution data never becomes a fake  FILLED  status After the approved Agent OS action returns, DeskFlow builds an execution receipt from the returned values, not from estimates: • venue • symbol • filled quantity • execution price • fee • decision mid • implementation shortfall • final status • Agentic account scope Fees remain separate from shortfall, so the audit does not hide transaction costs inside a flattering number. Every accepted result is appended to the blotter. Every rejected request is recorded in the rejection log. The result is a trace from natural-language intent to governed ticket to confirmed execution to post-trade evidence. DeskFlow exposes the workflow through MCP tools: > deskflow_price_order > deskflow_confirm_execution > deskflow_update_mandate > deskflow_inspect_blotter > deskflow_verify_mandate Install and connect it with Claude Code: > pip install git+github.com/Datwebguy/DeskFlo… > claude mcp add --scope user binance-mcp-server --transport http agent.binance.com/mcp/agenti… > claude mcp add --scope user deskflow -- deskflow mcp Live product page: datwebguy.github.io/DeskFlow… Source code: github.com/Datwebguy/DeskFlo… DeskFlow is not an AI trader trying to sound confident. It is an execution boundary designed to make confidence prove itself #Binance #AgentOS #MCP #AIAgents #BuildInPublic

been building SignalGate for @Telegraphprotoc Hackathon Season I, Track 3 agents sign blind way too often > stale RPCs, model guesses, one bad tx and a wallet's drained SignalGate sits in front of that moment: before any action fires, it pays live Telegraph miners in parallel to score it, and nothing proceeds until they agree fail-closed by design clean consensus → ALLOW with receipts real risk → BLOCK, wallet auto-quarantined miners disagree or confidence is low → WAIT every ask is a real x402 payment on Base Sepolia, so every verdict has an explorer link you can check yourself > watchlist keeps flagged wallets monitored 24/7, > stream shows live miner signals coming in real-time > halt ledger logs every quarantine and release zero key custody, which proves SignalGate never signs, it only advises try it live: signalgate-one.vercel.app