Lookonchain APP

App Store

OpenCode launches OpenTunnel: Mobile devices can now remotely use AI tools stored on personal computers.

1 days ago

Beating AI Express News: OpenCode co-founder Dax Raad has launched OpenTunnel, an open-source internal network penetration tool that generates public HTTPS links for websites, APIs, or OpenCode services running on a local computer. For example, users can run OpenCode on their home PC and remotely connect via a mobile device when away to continue managing AI programming tasks, without needing to configure a public IP or router port forwarding. This functionality is similar to Tailscale Funnel, both enabling local services to be exposed to the public internet without requiring visitors to join a private network. Both tools also support end-to-end encryption: traffic is not decrypted until it reaches the local device, so intermediate servers cannot read normal communication content. OpenTunnel also provides an SDK, allowing developers to integrate public network access features directly into their own applications. OpenCode has completed integration with OpenTunnel and plans to launch the remote pairing feature in its official version on October 9. At that time, users will be able to connect to OpenCode running on their computer via a mobile device using a one-time link or QR code. However, public links are not private connections—anyone can attempt to access the public URL, so sensitive services still require authentication. The community also notes that the operator holds domain and certificate issuance privileges, meaning they could theoretically issue additional certificates. Dax acknowledged that users still need to trust the operator, adding that users can deploy their own intermediate servers.

Relevant content

Analysis: AI leading stocks pushed the S&P 500 index up 117% during the bull market, with mounting market concentration risk.

Market analysis notes that since October 2022, the S&P 500 index has risen 117% cumulatively, driving the total market capitalization of U.S. stocks to increase by nearly $40 trillion, with the rally mainly fueled by AI-related stocks. Meanwhile, the S&P 500 Equal Weight Index has underperformed its market-cap weighted counterpart by 52 percentage points, hitting the widest gap in history—underscoring this bull market’s reliance on a small number of large tech stocks. Nvidia’s share price has surged more than 1900% cumulatively over the same period, while U.S. Treasury yields have recently risen to 5.34%, further stoking market concerns about valuation pressures. Investors are closely monitoring whether corporate earnings growth can sustain this bull market.

9 minutes ago

Injective Releases INJ and Proposed ETF Operational Guidelines, Detailing Token Economic Model and On-Chain Financial Applications

Injective has published a post explaining the relationship between its blockchain network, native token INJ, and the proposed Canary Staked INJ ETF (ticker: INJC). The blockchain serves as a transaction settlement infrastructure for financial applications, while INJ is used to pay network fees, stake to maintain network security, and participate in governance. If approved, INJC will offer investors exposure to INJ price movements and staking yields via traditional securities accounts. On the token economic front, Injective distributes staking rewards through protocol issuance and reduces circulating supply via token burns. Its Community BuyBack mechanism ties ecosystem revenue to INJ burns: participants receive a proportional share of an ecosystem asset basket based on the percentage of INJ they commit, while their committed INJ is burned. Regarding the ETF, Canary Capital has filed a related application with the U.S. SEC, naming BitGo Bank & Trust as the proposed custodian for INJ. Per the application documents, the fund plans to stake at least 90% of its INJ under normal conditions, with staking rewards typically reinvested. The product remains a proposal and is pending regulatory approval.

9 minutes ago

Iran's Revolutionary Guard: Will henceforth punish vessels violating rules outside the Strait of Hormuz.

Iran’s Islamic Revolutionary Guard Corps (IRGC) said it attacked a liquefied petroleum gas (LPG) tanker in the Strait of Hormuz several hours ago. The group later released a statement noting that effective immediately, any vessel committing violations outside the Strait of Hormuz will also face "punishment". The statement did not elaborate further on the specific scope of the violations or the details of the punishment measures.

9 minutes ago

Circle Issues Additional 250 Million USDC on Solana Network

According to Whale Alert monitoring, a Circle address minted 250 million USDC on the Solana blockchain five minutes ago, valued at approximately $250 million.

9 minutes ago

Ledger users suspected victims of large-scale crypto theft attack; on-chain analysts report losses exceeding $86 million

On-chain analyst Specter published a report noting that multiple Ledger user wallet theft incidents have been recently reported on X and Reddit. After tracing the relevant stolen addresses, he discovered the funds came from hundreds of victim wallets, spanning multiple mainstream blockchains including Ethereum, Tron, and Bitcoin, with total losses exceeding $86 million. The analyst has now disclosed multiple addresses linked to the cases, though the attack’s cause, exact number of affected users, and whether Ledger has security vulnerabilities remain unconfirmed. Current information is primarily sourced from on-chain tracking and community reports and requires further verification.

9 minutes ago

An iPhone version of Doubao mobile assistant has emerged? Developer demonstrates controlling an iPhone using Opus 5.5

Beating AI Express: Developer Finn Voorhees has demonstrated an AI agent capable of directly operating an iPhone. He only sent one prompt asking the agent to find the cheapest Ryanair flight to London for the next day. The agent then opened Ryanair on its own, input the destination, selected the date, checked fares, and finally returned to the chat interface to report the result—no computer was involved throughout the process. This agent runs in PiChat, an iPhone chat app developed by Voorhees. It leverages his previously open-sourced PiDurableKit at its core, running the Pi Agent on the iPhone via Apple’s JavaScriptCore to handle model calls, tool execution, and task management. The video uses Claude Opus 5.5, with model inference still running in the cloud. To enable the agent to continue working after switching to other apps, PiChat also uses Apple’s background task mechanism and saves execution progress via SQLite for easy resumption after interruptions. The most critical cross-app control code has not been made public. When activating iPhone Control in the video, the system required entering the lock screen password to enable XCTest UI automation. XCTest is Apple’s UI testing framework that can simulate taps, swipes, and text input. The open-source project TouchSynthesis leverages XCTest’s private interfaces to let iPhones generate touch events on their own, eliminating the need for real-time computer control. However, Voorhees has not confirmed whether he used this approach.

9 minutes ago

Popular tokens

BitcoinEthereumHyperliquidSolanaTRONBNBTetherAaveXRPPepeFartcoinOndoJupiterUniswapBonkPendleEthenaArbitrumAvalancheLidoChainlinkPolygonDogecoinCardano