Neglect of the modern Invoice (withsahel.com)
The article argues that businesses should treat modern invoicing as a designed communication channel—improving clarity, timeliness, and value signaling—to reduce friction and strengthen customer relationships.
Larger and more instructable language models become less reliable (pmc.ncbi.nlm.nih.gov) AI
The article reports that as large language models have been scaled up and “shaped” with instruction tuning and human feedback, they have become less reliably aligned with human expectations. In particular, models increasingly produce plausible-sounding but wrong answers, including on difficult questions that human supervisors may miss, even though the models show improved stability to minor rephrasings. The authors argue that AI design needs a stronger focus on predictable error behavior, especially for high-stakes use.
Ex-Meta worker investigated for downloading 30k private Facebook photos (bbc.com)
A former Meta employee in London is being investigated by the Metropolitan Police after he was suspected of downloading around 30,000 private photos from Facebook users by using a program to bypass security checks. Meta says it discovered the breach over a year ago, fired the engineer, and referred the case to law enforcement, leading to an arrest in November 2025. The police say the investigation is being handled by their Cybercrime Unit after a referral from the FBI, and Meta says it notified affected users and upgraded its security.
Bill Phillips used flowing water to model the economy (npr.org)
The piece recounts how economist Bill Phillips used analogies and models based on flowing water to explain how economic systems behave and to help visualize relationships between inflation, unemployment, and growth. It traces the approach’s origins and why the “water” metaphor proved useful for thinking about complex economic dynamics.
We need re-learn what AI agent development tools are in 2026 (blog.n8n.io) AI
The article argues that by 2026 many core “AI agent builder” capabilities—like document grounding, evaluations integrations, and built-in web/file/tool features—have become table stakes via mainstream LLM products. It proposes updating agent development evaluation frameworks to focus more on enterprise-readiness (security, observability, access controls, sandboxing, reliability) and on how agents can operate deterministically within controlled workflows while still allowing safe autonomy like spawning sub-agents. The author also notes shifting emphasis away from MCP-style interoperability after security concerns, and suggests reassessing how coding agents should be evaluated versus their role inside broader automation pipelines.
'This Is Just Not How the Human Race Should Operate,' Says US Senator (commondreams.org)
U.S. Sen. Chris Murphy condemned President Donald Trump’s threat to destroy Iran’s “whole civilization” if it does not reopen the Strait of Hormuz, calling it “pure evil,” a war crime, and fundamentally wrong even if Iran agrees to a deal. The article notes widespread condemnation from other lawmakers and human-rights advocates, with some calling for Trump’s removal from office and warning that the rhetoric could normalize extreme escalation.
ICE acknowledges it is using powerful spyware (text.npr.org)
ICE has acknowledged it uses spyware tools, including the “Graphite” technology, to intercept encrypted communications as part of efforts against fentanyl traffickers. The disclosure, made in a letter dated April 1 by ICE acting director Todd Lyons, follows an earlier congressional inquiry and expands scrutiny of how such tools could be used inside the United States. Civil liberties groups questioned the legal oversight and potential for misuse, while DHS said the technology is used to support law enforcement investigations consistent with privacy interests.
RSoC 2026: A new CPU scheduler for Redox OS (redox-os.org)
Redox OS is updating its CPU process scheduler for RSoC 2026 by replacing the legacy Round Robin algorithm with a Deficit Weighted Round Robin (DWRR) scheduler that can prioritize different process contexts. The post describes how the scheduler assigns weighted “tokens” to multiple queues, then uses an interleaved variant to reduce starvation and improve responsiveness under heavy CPU load. Benchmarks in an isolated test harness and in the pixelcannon demo are reported as showing significant responsiveness and throughput gains, especially compared to the old Round Robin approach.
When the compiler lies: breaking memory safety in safe Go (ciolek.dev)
Jakub Ciolek reports two Go compiler bugs (CVE-2026-27143 and CVE-2026-27144) that can break Go’s memory-safety guarantees using only safe Go code. The issues involve the compiler making incorrect “proved safe” assumptions—one around signed integer wrap in induction-variable bounds leading to missing bounds checks, and another around overlapping-memory handling being skipped due to a no-op conversion. Ciolek notes the common failure mode is “counterfeit certainty” across different compiler phases, stressing that memory safety depends on the whole toolchain, not just the language.
USD Purchasing Power in Real Time Since 2000 (onedollar.today)
The article tracks how the real purchasing power of $1 has eroded since January 2000, using CPI-U data and interpolating the trend forward in real time.
AI Assistance Reduces Persistence and Hurts Independent Performance (arxiv.org) AI
A paper on arXiv reports results from randomized trials (N=1,222) showing that brief AI help can reduce people’s persistence and impair how well they perform when working without assistance. Across tasks like math reasoning and reading comprehension, participants who used AI performed better in the short term but were more likely to give up and did worse afterward without the system. The authors argue that expecting immediate answers from AI may limit the experience of working through difficulty, suggesting AI design should emphasize long-term learning scaffolds, not just instant responses.
Show HN: Unicode Steganography (steganography.patrickvuscan.com)
The post demonstrates two forms of Unicode steganography that hide messages in plain text—using zero-width characters or homoglyph substitutions—then compares their detectability, capacity, and robustness. It argues that while current scanners and platform filters can catch some methods, a sufficiently capable AI could potentially generate encodings that evade human review and automated checks, raising practical concerns for LLM safety.
Move Detroit (movedetroit.com)
Move Detroit’s “Make Detroit Home” program is accepting applications to offer retention and relocation benefits to current, returning, and new Detroiters. The program plans to award more than $500,000 in benefits to 313 people, including local discounts and professional/community programming, with a limited $15,000 stipend for some applicants and $1,000 relocation support for others. It is aimed at Detroit creatives, entrepreneurs, and small business owners, while also seeking to welcome former and remote-work or critical-experience newcomers back to the city.
The Clock (blog.senko.net)
The article describes a “no-numerals” style timekeeping device built as a set of rotating rings that use physical anchors like Earth’s rotation around the Sun and local geography rather than traditional clock markings. Starting from a day/night cycle and expanding inward to show hours, minutes, and seconds (along with DST effects), it later adds an orbit-based outer ring for a stylized calendar. The author concludes that while the design can be made intuitive for potential outsiders (aliens), it still ultimately relies on human concepts like seconds and human-friendly time divisions.
Bitcoin and quantum computing (nehanarula.org)
The article argues that a cryptographically-relevant quantum computer would eventually break Bitcoin’s signature scheme, making it an existential risk unless Bitcoin is upgraded to post-quantum cryptography well before such systems arrive. It discusses uncertainty about both the likelihood/timeline of a quantum capability and Bitcoin’s ability to coordinate a consensus-and-wallet upgrade, noting that past protocol upgrades took years. The author concludes that, regardless of price speculation, the complexity and costs of switching should be started now because the upgrade may be difficult to complete in time.
A whole boss fight in 256 bytes (hellmood.111mb.de)
The article describes “Endbot,” a 256-byte DOS demo that runs in DosBox-X and produces a complete audio-visual scene: a robot sprite with accumulating bullet damage, a mathematically defined expanding explosion, and a scrolling checkerboard landscape, all while streaming a MIDI soundtrack from the demo’s own tiny code/data block.
A database of analog cameras that can be 3D printed (printed.analogcamera.space)
The site printed.analogcamera.space compiles a searchable database of analog camera designs that can be 3D printed, with options for different lens types, film formats, and custom build categories. It includes filterable listings for formats ranging from 35mm and 120 to larger formats, plus support for lens and camera type variations.
Xilem – An experimental Rust native UI framework (github.com)
Xilem is an experimental Rust native UI framework from linebender that uses a React/SwiftUI/Elm-style reactive approach to keep a lightweight view tree and update the rendered UI as that tree changes. It’s paired with the Masonry crate, which provides a retained widget tree and handles event/update passes, while Xilem offers the higher-level reactive programming model. The project runs on common Rust GUI and graphics libraries (winit for windows, wgpu/Vello for 2D rendering, and Parley/Fontique for text) and includes a web backend as well.
JSIR: A High-Level IR for JavaScript (discourse.llvm.org)
The LLVM MLIR community proposal describes JSIR, a high-level intermediate representation for JavaScript that preserves AST information to enable lossless round-trips between source, AST, and JSIR. Using MLIR regions for control flow and built-in dataflow analysis support, JSIR is presented as a foundation for source-to-source transformations such as decompilation, deobfuscation, and other tooling, and reports that it is already used in production at Google. The RFC emphasizes a near 1-to-1 mapping from ESTree nodes to JSIR operations and discusses design choices like reversible lowering back to AST.
Show HN: An interactive map of Tolkien's Middle-earth (middle-earth-interactive-map.web.app)
The article introduces “Middle-earth — Interactive Map,” a fan/educational web app that lets users explore Tolkien’s world with an interactive map, timeline, and clickable markers for events across the legendarium. It includes features such as filtering by book, toggling journey paths, and measuring distances between points, along with a note that the base map artwork is sourced from the internet for fan use.
Binary obfuscation used in AAA Games (blog.farzon.org)
The post summarizes a talk at Thotcon about how AAA game studios use binary obfuscation techniques and argues these methods can be designed to avoid breaking compiler optimizations like LTO.
US and Iran agree to provisional ceasefire (theguardian.com)
The US and Iran agreed to a two-week conditional ceasefire after a last-minute diplomatic push by Pakistan, with Tehran saying it will reopen the Strait of Hormuz. Donald Trump had set an imminent deadline to bomb Iranian power plants and bridges, but suspended the planned attacks if Iran provides “safe opening” for passage. Iran’s foreign minister welcomed the truce, while the article notes continued concerns about earlier strikes on infrastructure and international law.
Tailslayer: Library for reducing tail latency in RAM reads (github.com)
Tailslayer is an open-source C++ library designed to reduce tail latency in RAM reads by issuing “hedged” reads across multiple replicas stored on different DRAM channels with uncorrelated refresh behavior. It replicates data across channels using AMD/Intel/Graviton-compatible channel scrambling offsets, then returns the first completed replica while the library performs follow-up work on the winning value. The repo includes example code and supporting benchmarks/probes to characterize DRAM refresh timing.
Cells for NetBSD: kernel-enforced, jail-like isolation (netbsd-cells.petermann-digital.de)
Cells for NetBSD is an early-stage NetBSD-native system that provides lightweight, kernel-enforced, jail-like isolation for running multiple workloads on one host. It adds hardening profiles, supervised service execution, volume-backed persistence with built-in backup/restore, and centralized logging/metrics, managed via a host-side control plane (cellmgr) with an optional TUI (cellui). The project is aiming to bridge the gap between chroot and full virtualization while emphasizing explicit operational boundaries, with pre-release ISOs available for evaluation.