How Mark Klein told the EFF about Room 641A [book excerpt]
Community Discussion
The comments converge on a critical view of U.S. surveillance practices, emphasizing historical violations of the domestic‑foreign separation, ongoing covert data collection, and the difficulty of accountability. Many reference personal or documented experiences that illustrate lax security, secret infrastructure, and the impact on privacy, while also citing literature and legal debates surrounding FISA and Section 702. Sentiment is largely distrustful of government secrecy, supportive of transparency reforms, and skeptical of official assurances that domestic spying does not occur.
Opus 4.7 knows the real Kelsey
Summary
Claude Opus 4.7 can attribute unpublished text to its author with high accuracy when the author has a substantial public writing corpus. The tester pasted 125‑word political drafts, a school‑report excerpt, a movie review, fantasy prose, and an old college essay; Opus 4.7 consistently identified the writer as Kelsey Piper, while other models (ChatGPT, Gemini) gave different guesses. The model’s justifications are often incoherent, suggesting it detects subtle stylistic “tics” without true understanding. Deanonymization works best for writers with extensive real‑name publications; it fails on individuals lacking a sizable online footprint. However, even brief excerpts can link anonymous contributors to known acquaintances within a shared subculture. The author warns that as training data grows, the amount of text needed for accurate identification will shrink, threatening anonymity for prolific writers. Mitigation would require deliberately altering one’s style or using AI‑generated rewrites, both of which present practical challenges.
Read full article →
Community Discussion
Comments describe strong surprise at LLMs accurately attributing text to specific authors, noting that Opus 4.7 often succeeds where other models fail. The discussion converges on the view that such stylometric ability erodes online anonymity, with many expressing concern that personal writings can be traced despite pseudonyms or privacy tools. Some participants suggest countermeasures like local models or AI‑generated style masking, while others argue anonymity was never absolute. Overall sentiment blends intrigue about the technology’s capability with apprehension about its privacy implications.
For Linux kernel vulnerabilities, there is no heads-up to distributions
Summary
CVE‑2026‑31431 (“CopyFail”) is a Linux kernel local‑privilege‑escalation flaw introduced in kernel 4.14 (commit 72548b093ee38a6d4f2a19e6ef1948ae05c181f7). The issue was patched in the stable branches 6.18.22 (commit fafe0fa2995a0f7073c1c358d7d3145bcc9aedd8), 6.19.12 (commit ce42ee423e58dffa5ec03524054c9d8bfd4f6237) and upstream 7.0 (commit a664bf3d603dc3bdcf9ae47cc21e0daec706d7a5). Older long‑term kernels—6.12, 6.6, 6.1, 5.15, 5.10—remain unpatched; backporting is complicated by API changes. An attempted backport failed, and a temporary mitigation was provided as a patch disabling the “authencesn” crypto module (attached as 0001‑crypto‑disable‑authencesn‑module‑for‑CVE‑2026‑31431.patch). The author notes that kernel vulnerability notices are not automatically forwarded to distribution mailing lists unless the reporter does so.
Read full article →
Community Discussion
The comments criticize the current Linux kernel vulnerability disclosure flow, arguing that relying on individual reporters to notify downstream distributions is inadequate and that the kernel security team should assume that responsibility. Many express concern over premature public release of the exploit and its impact on shared‑hosting providers, while also noting that several distributions have already applied patches or can mitigate the issue with workarounds such as eBPF filters or default nosuid mounts. Opinions acknowledge that not every system is affected, but consensus calls for improved coordination and stronger default security settings.
Shai-Hulud Themed Malware Found in the PyTorch Lightning AI Training Library
Summary
The PyPI package **lightning** (versions 2.6.2 and 2.6.3, published 2026‑04‑30) was compromised in a supply‑chain attack. The malicious releases contain a hidden `_runtime` directory with an obfuscated 14.8 MB JavaScript payload that executes automatically on module import. The payload:
* Exfiltrates GitHub, npm, AWS, Azure, and GCP credentials, environment variables, CI/CD secrets, and local token files (e.g., `ghp_`, `gho_`, `npm_`).
* Sends data via four parallel channels: HTTPS POST to an encrypted C2 URL, GitHub commit‑search dead‑drop (prefix `EveryBoiWeBuildIsAWormyBoi`), attacker‑controlled public GitHub repos (Dune‑themed names), and direct pushes to the victim’s own repository when a `ghs_` token is present.
* Propagates to npm by injecting a `setup.mjs` dropper and `router_runtime.js` into packages the compromised npm token can publish, setting `scripts.preinstall` to execute the dropper.
Persistence is achieved through developer‑tool hooks: a Claude Code `settings.json` hook and a VS Code `tasks.json` task that run `setup.mjs` (installing Bun if needed). The malware may also add a malicious GitHub Actions workflow to dump repository secrets.
Indicators of compromise include commit messages prefixed with `EveryBoiWeBuildIsAWormyBoi` and public repos described “A Mini Shai‑Hulud has Appeared”. Remediation requires auditing for the listed files, rotating all discovered tokens, and rescanning projects with Semgrep rules.
Read full article →
Community Discussion
Comments express growing alarm over an apparent rise in high‑profile supply‑chain compromises, especially in Python packages, and criticize the scarcity of user‑friendly security tooling for hobbyist and machine‑learning projects. Contributors note that many dependencies are introduced without adequate auditing, question the trustworthiness of alternative Python installers, and advocate for stricter dependency pinning, sandboxing, or even self‑contained builds. While some point to recent mitigations such as pip cooldowns and locked lockfiles, the overall view is that the ecosystem remains vulnerable and in need of better detection and preventive measures.
I Got Sick of Remembering Port Numbers
Summary
The post describes “local.vibe,” a macOS‑only, open‑source tool (MIT licensed) that eliminates the need to remember local development port numbers. Implemented as a single Go binary, it installs dnsmasq, a pf port‑forwarding rule, and a trusted local CA, then runs as a daemon communicating via a Unix‑socket reverse proxy. For each project, a `vibe.json` defines a name and command; when started, local.vibe auto‑assigns an unused port, injects it as `$PORT`, and creates a `.vibe` hostname (e.g., `https://blog.vibe`) that proxies to the service over HTTPS. A dashboard at `https://local.vibe` lists running services, supports start/stop, editing, emoji icons, and grid or list views. It also provides built‑in proxies such as `tailscale.vibe` and `hass.vibe`. An HTTP API (e.g., `http://localhost:7999/setup.md`) enables AI agents to query or control the environment. Installation is a one‑line `setup.sh` after cloning the repository.
Read full article →
Community Discussion
The discussion centers on practical methods for managing service ports, with users sharing personal setups that combine configuration files, custom scripts, and tools like outport.dev and a GitHub “sonar” utility. Participants note advantages of automatic start‑stop mechanisms and resolver tricks, while also highlighting drawbacks such as AI agents mishandling process lifecycles and default port choices that feel inconvenient. Overall, the comments reflect a constructive exchange of alternatives, a desire for clearer port naming, and mild frustration with existing defaults and automation shortcomings.
Can I disable all data collection from my vehicle?
Summary
The Rivian Support Center page includes a “Reserving and configuring” section, which provides guidance for customers on how to reserve a vehicle and set up its configuration options. The page also features visual content presented as four images, each labeled with descriptive alt text: “Purchasing,” indicating steps or information related to buying a Rivian; “Products,” highlighting the range of vehicles and accessories; “Ownership,” covering topics such as maintenance, warranties, and user responsibilities; and “Company,” likely referencing corporate information or brand identity. These elements are organized to assist users in navigating the reservation process and understanding key aspects of Rivian’s product lineup and ownership experience.
Read full article →
Community Discussion
Comments show broad concern that modern EVs rely heavily on always‑on connectivity, raising privacy, security and recall‑remedy questions, especially where OTA updates lack a non‑digital fallback. Many criticize the need for dealer appointments to disable eSIMs outside Canada, calling for simple in‑vehicle toggles or physical disconnects, and note that turning off data often disables safety features like lane‑keeping, which they view as a dark‑pattern. While a minority praise the availability of a privacy option, the prevailing sentiment is skeptical, urging stronger regulations, clearer opt‑outs, and hardware‑level solutions.
Maladaptive Frugality
Summary
The author describes “maladaptive frugality,” a pattern where excessive cost‑avoidance leads to missed opportunities and unnecessary stress. A personal example involves paying for an iPhone repair without realizing AppleCare would have covered it, illustrating how hesitation over a modest expense can hinder larger projects. The behavior traces back to childhood lessons that framed frugality as a moral virtue and spending as sinful, a mindset reinforced by parents who emigrated from Hong Kong—a culture where historical economic instability fosters deep‑rooted saving habits. While such prudence can yield benefits like low‑cost travel, it also creates procrastination on essential purchases, resulting in resource shortages. The author recommends recognizing this bias, shifting focus to present‑moment decisions, and treating frugality as a tool rather than a controlling principle, especially for high‑impact choices. This mindfulness aims to balance cost‑consciousness with strategic investment in quality‑of‑life improvements.
Read full article →
Community Discussion
The comments discuss frugality as a culturally rooted habit, noting its persistence in post‑Soviet societies and its contrast with pervasive debt‑driven consumption in the United States. Many describe personal satisfaction from careful saving and the virtue it can represent, while also highlighting the downsides when it becomes rigid, creates emotional strain, or hampers enjoyment. Partners are cited as useful mirrors for balancing spending habits. Overall, the consensus emphasizes the need for moderation, recognizing both the benefits of mindful saving and the risks of excessive restraint.
CPanel and WHM Authentication Bypass – CVE-2026-41940
Summary
- **Scope**: CVE‑2026‑41940 is an authentication‑bypass flaw in all currently supported cPanel & WHM versions. It resides in the session handling code (Cpanel/Session.pm, Session/Load.pm, Session/Encoder.pm).
- **Root cause**: `saveSession` fails to strip CR / LF characters and only encrypts the `pass` field when a per‑session secret (``) is present. When the cookie omits the `` segment, the encoder is skipped, allowing a raw password containing `\r\n` to be written directly to the on‑disk session file. The injected line breaks create new top‑level entries (e.g., `hasroot=1`, `tfa_verified=1`, `cp_security_token=…`).
- **Exploitation**: An attacker can (1) trigger a pre‑auth session via a failed login, (2) resend the cookie without the `` part, and (3) send a Basic‑auth header whose password includes `\r\n` payload. The crafted session file is then used by cPanel to grant elevated privileges.
- **Patch**: Updated releases (110.0.x → 11.110.0.97, 118.0.x → 11.118.0.63, 126.0.x → 11.126.0.54, 132.0.x → 11.132.0.29, 134.0.x → 11.134.0.20, 136.0.x → 11.136.0.5) add a mandatory `filter_sessiondata` call inside `saveSession` and enforce `` validation.
- **Mitigation**: Upgrade to the patched versions immediately; apply network‑edge active‑defense rules to block malformed Basic‑auth headers and ensure session cookies include the `` component.
Read full article →
Community Discussion
The comments collectively emphasize the risks of custom‑written security components, urging reliance on established, widely‑audited solutions for session handling, authentication, and encryption. They criticize the cPanel vulnerability as an illustration of outdated or poorly maintained code, noting its broad impact on many sites, especially those using WordPress on cPanel. Concerns are raised about the difficulty of patching such flaws and the exposure of personal or homelab systems, leading to a call for stricter software standards and more rigorous update practices.
I built a Game Boy emulator in F#
Summary
A software engineer built “Fame Boy,” a functional‑style Game Boy emulator in F# to learn computer hardware. After completing a NAND‑to‑Tetris course and a CHIP‑8 emulator, he designed a core that interacts with desktop and web front‑ends via four simple elements: a 160×144 framebuffer, a 32 768 Hz ring audio buffer, a stepEmulator() function that executes one CPU instruction and returns cycles, and a getJoypadState callback. The Sharp LR35902 CPU is modeled with pure F# discriminated unions, reducing 512 opcodes to 58 typed instructions and preventing illegal states at compile time. Memory.fs provides the RAM and bus; IoController.fs abstracts hardware registers, while the stepper function sequentially runs CPU, timers, serial, APU (4× CPU cycles), and PPU (4× CPU cycles) to emulate the parallel hardware on a single thread. Mutability is used for performance‑critical arrays. Unit tests—generated with AI prompts—drive test‑driven development. The PPU renders whole scanlines instead of a pixel FIFO, simplifying implementation but missing timing‑sensitive effects. Joypad handling required read‑on‑demand updates to emulate the register’s dual‑read behavior. An APU was added later, with timing driven by frame rate rather than audio sampling, highlighting challenges in accurate audio emulation.
Read full article →
Community Discussion
The comments express strong enthusiasm for the F# emulator, praising the functional approach, the project’s educational value, and its inspirational effect on related ideas. Several contributors offer technical suggestions, such as using struct discriminated unions and simplifying byte setters, while others note performance limits and the dominance of C#‑centric libraries in the F# ecosystem. Interest in visual debugging tools, unique emulator features, and cross‑platform experiments is evident, alongside brief skepticism about AI involvement and occasional off‑topic remarks. Overall sentiment is positive and constructive.
Claude Code refuses requests or charges extra if your commits mention "OpenClaw"
Summary
The page displays a generic error notice indicating that an operation failed and advises the user to retry (“Something went wrong, but don’t fret — let’s give it another shot”). No additional context, instructions, or content is provided beyond this brief message. The only visual element referenced is a single image, identified solely by its alternative text, which consists of a warning emoji (⚠️). No further description, caption, or functional information about the image is included. Consequently, the page offers no substantive information, data, or technical details beyond the error prompt and the placeholder image reference.
Read full article →
Community Discussion
The comments convey widespread frustration with Anthropic’s recent handling of OpenClaw‑related requests, sudden usage limits, and opaque anti‑abuse measures that trigger unexpected billing. Users criticize the apparent reliance on simple keyword bans, inconsistent enforcement, and perceived leadership missteps, noting a loss of trust and growing interest in alternative providers such as OpenCode, DeepSeek and self‑hosted models. While a minority suggest separate pricing tiers or acknowledge capacity constraints, the dominant view is that the current approach is poorly communicated, harmful to customers, and prompting many to consider switching services.