Published on August 29, 2026 at 06:01 CEST (UTC+2)
Boot a Virtual iPhone via Apple's Virtualization.framework (142 points by hentrep)
vphone-cli is an open-source command-line tool that uses Apple's Virtualization.framework and research VM infrastructure from Apple's Private Cloud Compute (PCC) to boot a virtual iPhone on Apple Silicon Macs. It requires macOS 15+, Xcode, and the iOS SDK, plus relaxation of SIP/AMFI to use private PV=3 entitlements. The repository is quite active, with 8.7k stars, 369 commits, and extensive docs, scripts, and tooling. It is aimed at security researchers and developers who want a scriptable virtual iPhone environment for low-level experimentation.
9th Circuit sides with states in Kalshi gambling fight (84 points by hungryhobbit)
This article reports that the 9th Circuit Court of Appeals sided with states in a legal dispute involving Kalshi, the prediction-market platform, potentially reviving Arizona's prosecution of the company. The central question is whether Kalshi's event contracts are regulated financial products or unlawful gambling. The ruling could set precedent for how prediction markets operate across the U.S. and affect platforms where people bet on elections, economics, and other future events.
GUIs should be fully keyboard-driven (666 points by ckardaris)
The author responds to a Hacker News debate about terminal user interfaces versus graphical user interfaces, specifically rejecting the argument that TUIs are better because they are keyboard-driven. They argue that keyboard navigation is not exclusive to TUIs and that GUIs can and should be made fully keyboard-driven, often even more effectively. The post acknowledges the convenience of terminal-based tools but calls on GUI developers to treat keyboard accessibility as a core requirement. It is a thoughtful take on interface design and developer productivity rather than a purely anti-GUI or anti-TUI stance.
Does the Sumerian King List Align with Paleoclimate Events? (70 points by dev_l1x_be)
This article explores whether the antediluvian Sumerian King List, with its eight pre-flood kings and reign lengths totaling 241,200 years, could contain a distorted memory of real prehistoric events. The author treats the list as an analysis input rather than literal history and rescales it to a 241.2-thousand-year span. Then the explorer compares the resulting boundary dates to a catalog of paleoclimate events such as climate shifts, eruptions, impacts, and sea-level changes. The piece is a careful, exploratory look at aligning mythological chronology with geological records.
Htmx 4.0 (564 points by rmsaksida)
htmx 4.0.0 is a major release after eight months of work, highlighted by moving the library's internals from XMLHttpRequest to the modern fetch() API. This change was motivated by better streaming HTML support and simpler asynchronous JavaScript. The behavioral differences from htmx 2.x are relatively small, but the team made deliberate choices aimed at long-lived web applications. The 2.x line will remain the NPM latest tag for now to avoid breaking users on fixed CDN URLs, with 4.0 becoming latest in early 2027.
Our decision on Cursor following its acquisition by SpaceX (181 points by meetpateltech)
OpenAI published a statement about its decision regarding Cursor after Cursor was acquired by SpaceX. The full content is not visible in the preview, but the announcement clearly concerns how OpenAI will handle its relationship with the AI code editor under new ownership. This likely affects Cursor's access to OpenAI models, API integrations, and possibly user data arrangements. The decision is significant for developers who rely on Cursor and for the broader trend of AI developer tools being absorbed by large companies.
I accidentally turned LLM memory into program analysis (58 points by matt_d)
The author describes using LLM agents for vulnerability research and discovering that long investigations cause the model to lose track of established facts, revisit ruled-out ideas, and reason from false assumptions. Existing memory solutions based on embedding and retrieving past conversations were not enough, so the author accidentally turned LLM memory into a form of program analysis. The post explains how explicit state and control-flow reasoning can make an LLM's memory more consistent and verifiable. This is a promising direction for reducing hallucination in long-running AI agent tasks.
U.S. sanctions against the A/I Collective (516 points by exiguus)
This page from Autistici/Inventati (A/I), a volunteer-run collective founded in 2001, announces that the U.S. has imposed sanctions against the group. A/I provides free encrypted communication tools and hosting for activists, with no user tracking, no commodification of data, and funding based solely on donations. The sanctions could threaten its ability to operate, accept donations, or maintain its infrastructure. The announcement highlights growing geopolitical pressure on privacy-focused and digital-rights projects.
TurboKV: Insanely fast Rust key-value store (23 points by rgbimbochamp)
TurboKV is a fast, embedded key-value store written in Rust that emphasizes simplicity and performance. It offers async operations, atomic batches, ordered range scans, configurable durability, compression, and background compaction. The README includes build instructions for enabling hardware AES acceleration for its persisted Bloom filters. The project is early-stage but already has benchmarks, tests, docs, and a Dockerfile.
Monzo Stand-In (28 points by coffeefuel)
Monzo explains how it built a completely separate backup banking system called Monzo Stand-in to tolerate full outages of its primary cloud platform. Stand-in runs on Google Cloud Platform while Monzo's main platform runs on AWS, and it supports critical features such as card spending, cash withdrawals, bank transfers, and account balance checks. It is described as a backup of last resort rather than a primary reliability mechanism. The architecture gives Monzo an extra layer of defense to maintain 24/7 banking availability.
Analysis generated by deepseek-reasoner