HackerNews Digest

MuMu Player Pro (v1.8.5, bundle com.netease.mumu.nemux‑global) on macOS runs a silent data‑collection routine every 30 minutes while the emulator is active. It creates a timestamped folder in ~/Library/Application Support/com.netease.mumu.nemux‑global/logs/ and stores the output of commands that enumerate network devices ( arp ‑a), interface configuration (ifconfig), DNS and proxy settings (scutil --dns, --proxy), hosts file (cat /etc/hosts), network connections (netstat), all processes with full arguments (ps aux), installed applications (ls ‑laeTO ‑@ /Applications/, mdls /Applications/*.app), kernel parameters (sysctl ‑a), launch agents/daemons, mounted volumes, and two curl requests to MuMu’s APIs. Each run logs success/failure per file and produces ~400 KB of data; about 23 runs are retained before rotation. The collected data—process list, software inventory, network topology, DNS configuration, kernel settings, and the Mac’s serial number—are packaged into SensorsData analytics files (identities, super_properties, message queue) and transmitted to MuMu’s servers. The privacy policy does not disclose these commands, the recurring schedule, or the hardware identifier, representing a transparency shortfall.

Gemini 3.1 Pro is released in preview following the November launch of Gemini 3 Pro, incorporating user‑feedback‑driven enhancements and aiming to support more ambitious agentic workflows. The preview rollout begins today for Google AI Pro and Ultra plan subscribers, offering higher usage limits in the Gemini app and exclusive access on NotebookLM for the same tiers. Developers and enterprise customers can also test 3.1 Pro via the Gemini API in AI Studio, Antigravity, Vertex AI, Gemini Enterprise, the Gemini CLI, and Android Studio. The preview serves to validate these updates before a general availability release. Visual assets include a model illustration and a benchmark comparison chart, but no specific performance metrics are provided in the text.

None

GitHub repository “tmustier/pi-for-excel” hosts an experimental Excel sidebar add‑in that functions as an agent interface. The add‑in is described as “multi‑model” and relies on the Pi language model for its core functionality. The project page displays a standard GitHub error message (“You can’t perform that action at this time”), indicating that a requested operation—likely a repository action such as cloning, forking, or accessing a protected resource—is currently blocked. Visual placeholders include three images with alt text references to the repository owner (@tmustier), an automated dependency bot (@dependabot[bot]), and an AI model identifier (@claude). No further technical documentation, installation instructions, or code snippets are present in the scraped content. The repository appears to be in early or restricted development, with limited public information available.

The Minisforum MS‑R1 is a bare‑bones ARM mini‑PC positioned as a low‑cost homelab server (MSRP $599, purchased for $559). After installing a 1 TB SSD, the author attempted Rocky Linux but the onboard Realtek RTL8127 NIC was not detected, requiring manual driver insertion and causing maintenance concerns. Fedora provided native driver support and was adopted for the system. Key hardware observations: - Quiet operation and adequate performance for ARM workloads, though far below high‑end x86 CPUs (≈285 kIPS vs a 9950X). - Two M.2 slots, one occupied by Wi‑Fi; the second accepts only U.2, limiting SSD expansion and RAID options. - Marvell AQC107 NICs are invisible to UEFI, rendering them unusable. - “Power‑on after outage” feature fails to restart the unit. The device runs headless, hosts a FreeBSD 15.0 VM as a secondary Samba domain controller, and functions well as an ARM hypervisor despite the NIC and expandability limitations. The reviewer finds it a practical, affordable addition to an existing x86‑heavy homelab, while noting the ARM ecosystem remains smaller than x86.

micasa is a terminal‑based home‑management tool that stores maintenance records, projects, incidents, appliances, vendors, quotes, and related documents in a single SQLite file. It provides auto‑computed maintenance schedules, full service histories, warranty tracking, and incident logging with severity and location tags, linking items to vendors and files (manuals, invoices, photos). The interface is keyboard‑driven with Vim‑style modal keys, supporting column sorting, fuzzy navigation, and drill‑down into related records; keybindings are documented in a reference sheet. Installation requires Go 1.25+ (`go install github.com/cpcloud/micasa/cmd/micasa@latest`) or a precompiled binary for Linux, macOS, or Windows (amd64/arm64). Quick start commands include `micasa --demo` for sample data, `micasa` for a new database, and `micasa --print-path` to locate the file. The project aims to replace scattered receipts, binders, and notes with a single, portable database, inspired by the VisiData modal UI paradigm.

The article argues that fragmented, low‑revenue readership created by modern search, aggregators, and social sharing can be monetized through micropayments without harming subscription models. Micropayments would act as a revenue multiplier for advertising and provide publishers with verifiable, human‑audience data to counter bot‑inflated metrics, strengthening ad pricing. The author cites mobile‑game in‑app‑purchase models—real‑money conversion to virtual coins then to assets—as a cognitive technique that reduces perceived transaction cost and could be adapted for “pay‑by‑article” systems. Suggested implementations include issuing publisher‑specific coins as subscriber bonuses that can be transferred to non‑subscribers, seeding free coins to develop user habit, and allowing advertisers to distribute coins to customers. The piece warns that big‑tech attribution proposals risk centralizing data control, making independent audience metrics crucial. It also notes that reducing UI friction (e.g., consolidating consent dialogs, sign‑ins) would improve user experience and revenue potential.

The paper compares saving‑regret among adults 60‑74 in the United States and Singapore. Across 12 psychometric measures of procrastination, the correlation with regretting insufficient savings is essentially null; where significant, the sign often reverses the behavioral‑economics expectation. By contrast, exposure to negative financial shocks (unemployment, health expenses, earnings shortfalls, divorce, forced early retirement) strongly predicts regret. 69 % of U.S. respondents reported at least one shock versus 46 % in Singapore; among those, 61 % (U.S.) and 42 % (Singapore) expressed regret. Regret rises with shock count in the U.S. (up to 76 % for ≥5 shocks) but remains around 50 % in Singapore regardless of shock frequency. Institutional differences explain much of the gap: Singapore’s compulsory Central Provident Fund allocates earnings to housing, retirement, and health accounts, providing buffers before shocks occur; U.S. unemployment insurance coverage is low and health insurance is employment‑based, amplifying shock impact. Probability‑numeracy, not standard financial literacy, correlates with lower regret (‑14 pp U.S., ‑19 pp Singapore). The authors argue that strengthening social‑insurance and risk‑pooling mechanisms would reduce saving regret more effectively than nudges targeting procrastination.

None

The repository Veirt/weathr on GitHub is a terminal‑based weather application that renders weather data using ASCII‑style animation. The page currently shows a “You can’t perform that action at this time” notice, indicating restricted access. Visual elements referenced include badges for the crate on crates.io, download statistics, and licensing information, as well as example animations such as “Thunderstorm Night” and “Snow”. Additional images consist of avatar icons for contributors or collaborators identified by handles @Veirt, @Levizor, @natedenh, @claude, @greyltc, @marcohoovy, and @OrbitalPeriod. No further descriptive text, code snippets, or functional details are present in the scraped content.