Published on June 17, 2026 at 06:00 CEST (UTC+2)
GrapheneOS has been ported to Android 17 (497 points by Cider9986)
GrapheneOS has been ported to Android 17
GrapheneOS, the security-hardened Android distribution, has been fully ported to Android 17 and will release official builds imminently. The team tested the port on Pixel 6a, 7, 7a, 8, 10a, 10, and 10 Pro Fold, and is pushing code to public repositories. An initial Android 17 release is planned for the day after the announcement, with public testing to follow. The post clarifies that despite Google not supporting Android 17 for Pixel 6 and 6 Pro, GrapheneOS may still support those devices.
Running local models is good now (1090 points by jfb)
Running local models is good now
The author, a long-time local model user, reports that as of mid-2026, local models like GPT-OSS and Gemma 4 have become surprisingly capable. On a 2022 M2 Mac with 64 GB RAM, they now use local models for agentic coding loops at about 75% the accuracy/speed of frontier models. Local models have evolved from slow, inaccurate tools into fast, personalized development assistants that rarely require verification against API-based models.
SpaceX to buy Cursor for $60B (932 points by itsmarcelg)
SpaceX to buy Cursor for $60B
According to a Reuters report, SpaceX is acquiring Anysphere, the company behind the AI coding assistant Cursor, for $60 billion. This massive deal underscores the extraordinary valuation of AI developer tools and signals that major industrial corporations see AI coding assistants as critical infrastructure for engineering and software development.
Humiliating IIS servers for fun and jail time (134 points by denysvitali)
Humiliating IIS servers for fun and jail time
This detailed bug bounty guide walks through techniques to find and exploit misconfigured Microsoft IIS web servers. Methods include Shodan searches, Google dorking, tilde enumeration, and using LLMs for GitHub dorks to resolve shortnames. The author covers vulnerabilities such as internal IP disclosure, path traversal, web.config exposure, and authentication bypass, with an emphasis on automation tools like Nuclei.
Wolfram Language and Mathematica Version 15, AI Assistant, Symbolic Music, More (110 points by alok-g)
Wolfram Language and Mathematica Version 15, AI Assistant, Symbolic Music, More
Stephen Wolfram announces Version 15 of Wolfram Language and Mathematica, featuring a built-in AI assistant integrated into every notebook, along with symbolic music support, time series improvements, and gigabyte-sized notebook handling. The release positions AI as a core, seamlessly embedded capability rather than a separate tool, and introduces new functions for model fitting, categorical data, and multipanel visualization.
TIL: You can make HTTP requests without curl using Bash /dev/TCP (307 points by mrshu)
TIL: You can make HTTP requests without curl using Bash /dev/TCP
A short tutorial demonstrates how to use bash’s built-in /dev/tcp feature to make raw HTTP requests when curl, wget, or other tools are unavailable. By opening a file descriptor to a TCP socket and manually writing HTTP request lines, one can fetch endpoints from minimal containers. The method can include custom headers like Authorization tokens, making it a useful fallback for debugging restricted environments.
Calvin and Hobbes and the price of integrity (324 points by pseudolus)
Calvin and Hobbes and the price of integrity
This deeply-researched piece explores Bill Watterson’s unwavering commitment to artistic integrity, including his refusal to merchandise Calvin and Hobbes despite massive commercial pressure. It traces his early struggles at Kenyon College and his decision to vanish from the public eye rather than compromise his vision. The article is a meditation on creative freedom and the personal cost of principle.
GPT‑NL: a sovereign language model for the Netherlands (154 points by root-parent)
GPT‑NL: a sovereign language model for the Netherlands
TNO, SURF, and the Netherlands Forensic Institute are building GPT‑NL, an independent Dutch language model that prioritizes public values such as transparency, privacy, and sovereignty. The project aims to reduce dependence on foreign AI providers and ensure control over data and governance. It represents a broader European push for digital autonomy and responsible AI deployment.
Stop Using JWTs (290 points by dzonga)
Stop Using JWTs
This popular Gist argues that JSON Web Tokens (JWTs) should not be used for session management, citing security pitfalls, complexity, and unnecessary overhead. The author advocates for simpler opaque session tokens or other alternatives. The piece has sparked significant discussion about authentication best practices in modern web and AI application backends.
Has AI already killed self-help nonfiction books? (196 points by imakwana)
Has AI already killed self-help nonfiction books?
Tim Ferriss presents sales data showing a 26.3% year-over-year decline in self-help nonfiction units in Q1 2026, with his own book catalog experiencing similar drops. He attributes the trend to LLMs like ChatGPT and Claude that provide instant, customizable answers, replacing the need for how-to books. The article suggests a structural shift in how people consume instructional content, with AI acting as a direct competitor to the traditional publishing model.
1. Local models have crossed the “good enough” threshold for practical development tasks
- Trend: Models like Gemma 4 and GPT-OSS now run on consumer hardware (e.g., 64 GB RAM Macs) and achieve ~75% of frontier model performance for agentic coding. Developers report rarely needing to verify outputs against API models.
- Why it matters: This democratizes access to powerful AI, reduces reliance on expensive cloud APIs, and enables offline, privacy-preserving workflows. It also pressures API providers to differentiate on latency, cost, or unique capabilities.
- Implications: Expect an acceleration of on-device AI agents, edge deployments, and tooling that optimizes local model orchestration (e.g., llama.cpp, LM Studio). This may also shift the competitive landscape toward open-weight models.
2. Sovereign AI initiatives are gaining real institutional momentum
- Trend: GPT‑NL (Netherlands), alongside similar national projects in Europe and Asia, explicitly builds models for cultural and linguistic independence. The project emphasizes governance, transparency, and public values.
- Why it matters: Dependence on a handful of US/China-based AI providers raises concerns about data sovereignty, censorship, and alignment with local laws. Sovereign models offer a way for governments to control critical AI infrastructure.
- Implications: Expect more publicly funded or consortium-led LLMs tailored to specific languages and regulatory contexts. This could fragment the AI ecosystem and create new standards for model auditing and certification.
3. AI coding assistants are being valued as strategic industrial assets
- Trend: SpaceX’s reported $60B acquisition of Anysphere (Cursor) signals that AI coding tools are no longer just productivity aids but core engineering infrastructure for large-scale hardware and software projects.
- Why it matters: This validates the AI-assisted development paradigm as a cornerstone of competitive advantage. It also suggests that the next frontier is domain-specific coding assistants (e.g., for aerospace, automotive).
- Implications: Expect a wave of acquisitions in the dev-tool AI space, increased investment in proprietary fine-tuned models for internal codebases, and tighter integration between LLMs and CI/CD pipelines.
4. AI is rapidly cannibalizing traditional content markets, especially self-help and how-to nonfiction
- Trend: Tim Ferriss’s data shows a 26% drop in self-help book sales, attributed to LLMs providing instant, interactive answers. Other nonfiction subcategories also declined, while only hobby/craft and religion grew.
- Why it matters: This represents a fundamental shift in how people seek knowledge—from static, linear books to dynamic, conversational AI. Publishers face an existential threat unless they adapt (e.g., bundling AI companions with books).
- Implications: Content creators will need to pivot to “AI-native” formats (e.g., curated prompt sets, interactive guides). Expect a rise in AI-generated or AI-enhanced book services, and potential legal battles over training data derived from copyrighted works.
5. AI is being deeply embedded into existing professional software ecosystems
- Trend: Version 15 of Wolfram Language/Mathematica integrates a built-in AI assistant into every notebook, alongside new functionality like symbolic music and model fitting superfunctions. AI is no longer a bolt-on but a first-class primitive.
- Why it matters: This lowers the barrier to using AI for computational work (data analysis, simulation, symbolic reasoning) and blurs the line between “AI tool” and “application platform.”
- Implications: Expect other mature software suites (e.g., MATLAB, Excel, CAD tools) to follow suit. Users will demand natural-language interfaces for complex operations, and vendors that fail to embed AI risk obsolescence.
6. Security researchers are actively weaponizing LLMs for offensive operations
- Trend: The IIS hacking guide explicitly uses LLMs for GitHub dorks, shortname resolution via BigQuery, and automated fuzzing list generation. LLMs are becoming part of the penetration tester’s toolkit.
- Why it matters: AI accelerates vulnerability discovery and exploitation, widening the asymmetry between attackers and defenders. It also lowers the skill barrier for sophisticated reconnaissance.
- Implications: Security teams must adopt AI-powered defensive tools (e.g., LLMs for log analysis, automated patch validation). Organizations should expect AI-augmented bug bounties and more rapid disclosure of critical vulnerabilities. This also raises ethical questions around the release of AI-generated exploit code.
7. Privacy and authentication concerns remain unresolved in an AI-first world
- Trend: Despite advances in AI, fundamental security debates persist—GrapheneOS continues to prioritize OS-level hardening, and “Stop Using JWTs” argues for simpler session mechanisms. AI services often inherit these vulnerabilities (e.g., JWT-based API keys for LLM endpoints).
- Why it matters: As AI becomes embedded in sensitive workflows (healthcare, finance, government), robust authentication and OS security are prerequisites for adoption. High-profile breaches could derail trust in AI platforms.
- Implications: Expect stricter standards for AI API authentication, growing adoption of hardware-backed attestation (e.g., Android’s key store), and increased scrutiny of AI supply chain security. The tension between convenience (e.g., seamless Copilot access) and security will be a defining challenge for the next phase of AI deployment.
Analysis generated by deepseek-reasoner