2025-10-09 Hacker News Top Stories

# Tech & Policy Roundup — Privacy Wins, Policy Reversals, and New AI Models

---

## 🇩🇪 German Government Opposes EU's Warrantless Chat Surveillance

**Source:** [paddi_hansen on Twitter/X](https://xcancel.com/paddi_hansen/status/1975595307800142205)  
**HN Thread:** [1104 points | 501 comments](https://news.ycombinator.com/item?id=45506143)

Germany’s governing coalition (**CDU/CSU**) has taken a firm stance against the EU’s *anlasslose Chatkontrolle* — warrantless chat surveillance. This is seen as a rare and important win for digital privacy in Europe.

**Key points:**
- The government will reject proposals for communication monitoring without cause.
- Announcement delivered in a 40-second video by Daniel Eck.
- Widely supported by digital rights advocates.
- Calls for legal codification, oversight, and sunset clauses.
- Observers warn that political pressure could still shift policy.

> **Context:** Germany’s history with the Stasi and Gestapo informs deep public skepticism toward mass surveillance.

---

## 📉 Synology Reverses Third‑Party HDD Ban

**Source:** [Guru3D](https://www.guru3d.com/story/synology-reverses-policy-banning-thirdparty-hdds-after-nas-sales-plummet/)  
**HN Thread:** [981 points | 628 comments](https://news.ycombinator.com/item?id=45513485)

Synology has rolled back a 2025 policy limiting its NAS devices to first‑party drives. The move follows:
- A sharp sales drop.
- Strong user backlash and brand damage.
- Migration of loyal users to competitors.

With DSM 7.3, compatibility for third‑party HDDs and SATA SSDs is restored, without warnings or degraded performance.

**HN Commentary Highlights:**
- Leadership underestimated customer willingness to leave.
- Employee morale was damaged.
- Stock ownership plans are often PR tools with limited impact.
- Costco‑like “customer-first” models are more sustainable.

> **Lesson:** Platform lock‑in is risky; openness improves trust.

---

## 🤖 Google DeepMind’s Gemini 2.5 “Computer Use” Model

**Source:** [Google Blog](https://blog.google/technology/google-deepmind/gemini-computer-use-model/)  
**HN Thread:** [609 points | 312 comments](https://news.ycombinator.com/item?id=45507936)

**What it does:**
- Automates UI interactions via screenshots and commands.
- Works in a loop: screenshot → command → execute → repeat.
- Optimized for browsers (mobile UI also supported, desktop OS-level excluded).

**Demo Scenarios:**
- Pull pet data from a site directly into a CRM and schedule appointments.
- Organize Kanban boards by categorizing sticky notes.

**HN Insights:**
- Playwright + LLM script generation > direct MCP control for test automation.
- Best practice is human-navigation + script generation to minimize LLM calls.
- Herd platform offers reusable automation “trails”.

---

## 🐛 Bug in Go’s ARM64 Compiler

**Source:** [Cloudflare Blog](https://blog.cloudflare.com/how-we-found-a-bug-in-gos-arm64-compiler/)  
**HN Thread:** [485 points | 82 comments](https://news.ycombinator.com/item?id=45516000)

**Discovery:** Cloudflare uncovered a non‑atomic stack pointer modification issue in Go’s ARM64 compiler.

**Impact:**
- Fatal runtime panics (`traceback did not unwind completely`, SIGSEGVs).
- Triggered during stack unwinding in idle services.
- Root cause: incorrect handling of large immediate values in generated code.

**Fix:** Load constants into a temp register before affecting the stack pointer — one extra instruction, atomic change achieved.

---

## 📧 Grassroots Campaign Slows EU “Chat Control” Bill

**Source:** [Politico](https://www.politico.eu/article/one-man-spam-campaign-ravages-eu-chat-control-bill-fight-chat-control/)  
**HN Thread:** [388 points | 147 comments](https://news.ycombinator.com/item?id=45514433)

**Joachim** — a Danish software engineer — built “Fight Chat Control” over a single weekend.  
**Result:**
- 2.5M visits triggered millions of protest emails to EU lawmakers.
- Inspired debates in Denmark, Ireland, Poland.
- Signal threatened to exit EU if bill passes.

**Main concern:** Bill would force scanning of encrypted messages for CSAM — critics warn of mass surveillance risks.

---

## 🛠 Rails Simplicity vs. Modern Frontend Bloat

**Source:** [Doing Rails Wrong](https://www.bananacurvingmachine.com/articles/you-re-doing-rails-wrong)  
**HN Thread:** [365 points | 265 comments](https://news.ycombinator.com/item?id=45505692)

Conversation satire:
- Kevin advocates Vite + React + TypeScript + Tailwind + ESLint + Docker + CI/CD.
- John runs a single Rails command → app works instantly.
- Moral: you may not need complex frontend stacks; vanilla Rails still delivers.

---

## 📩 Public Agency’s “Open Source” Vendor Lock-in Horror

**Source:** [The Email They Shouldn’t Have Read](https://it-notes.dragas.net/2025/10/08/the-email-they-shouldnt-have-read/)  
**HN Thread:** [347 points | 145 comments](https://news.ycombinator.com/item?id=45515657)

Summary:
- Agency’s hosted “open-source” mail system secretly monitored client emails.
- Lock‑in via onerous contracts (long notice periods, arbitrary fees).
- Evidence collected via a planted “false” email.
- Lesson: Beware of “pseudo–open source” commercial packages and inspect contracts closely.

---

## 🚨 ICE’s Fake Cell Towers in Surveillance Vehicles

**Source:** [TechCrunch](https://techcrunch.com/2025/10/07/ice-bought-vehicles-equipped-with-fake-cell-towers-to-spy-on-phones/)  
**HN Thread:** [320 points | 195 comments](https://news.ycombinator.com/item?id=45505854)

- ICE spent $825K on vehicles with IMSI catchers (“Stingrays”).
- Capable of intercepting calls, messages, and tracking phones.
- Heavy use since 2013; oversight murky.
- Apple’s “Lockdown Mode” disables 2G/3G as part of high-security defenses.

---

## 🧠 Tiny Recursion Model (TRM) – Small Network, Big Reasoning

**Source:** [alexiajm.github.io](https://alexiajm.github.io/2025/09/29/tiny_recursive_models.html)  
**HN Thread:** [295 points | 60 comments](https://news.ycombinator.com/item?id=45506268)

- 7M parameter recursive model achieves surprising AGI benchmarks (ARC-AGI-1: 45%).
- Works via iterative answer refinement.
- Challenges the “bigger is better” LLM narrative.
- Lower cost, reduced risk of overfitting.

---

## 🧪 Nobel Prize in Chemistry 2025 — MOF Pioneers

**Source:** [Nobelprize.org](https://www.nobelprize.org/prizes/chemistry/2025/popular-information/)  
**HN Thread:** [295 points | 52 comments](https://news.ycombinator.com/item?id=45514164)

Awarded to:
- **Susumu Kitagawa**
- **Richard Robson**
- **Omar M. Yaghi**

For creating **Metal–Organic Frameworks (MOFs)**:  
- Ultra-porous materials for drug delivery, CO₂ capture, hydrogen storage.
- Potential far exceeds zeolites, but industrial scaling is a challenge.
- Surface areas can reach **7,000 m²/g**.

---

## 💬 Closing Thought

Across these stories — from political resistance to mass surveillance, to corporate reversals on restrictive policies, to AI tools balancing complexity and efficiency — common threads emerge:

- **Openness builds trust** (Synology’s reversal, AiToEarn’s open model).
- **Simplicity can outperform complexity** (Rails vs bloated frontends, TRM vs huge LLMs).
- **Public participation matters** (Joachim’s campaign shifted EU politics).
- **Technical precision is vital** (Go ARM64 bug, Gemini UI automation).

---

**Resource Highlight:**  
Tools like **[AiToEarn](https://aitoearn.ai/)** embody these principles — allowing creators to generate AI-driven content, publish across Douyin, Kwai, WeChat, Bilibili, Xiaohongshu, Facebook, Instagram, LinkedIn, Threads, YouTube, Pinterest, and X simultaneously, with analytics and AI model rankings ([see rankings](https://rank.aitoearn.ai)) — supporting openness, efficiency, and user control.

---

Read more

Translate the following blog post title into English, concise and natural. Return plain text only without quotes.

ChatGPT Atlas 发布,AI 浏览器大乱斗...

Translate the following blog post title into English, concise and natural. Return plain text only without quotes. ChatGPT Atlas 发布,AI 浏览器大乱斗...

# AI Browsers: When LLM Companies Step In 原创 lencx · 2025-10-22 07:00 · 上海 --- ## Overview Large Language Model (LLM) companies are making moves into the **AI browser** space. From new entrants like **Dia**[1], **Comet**[2], and **ChatGPT Atlas**[3], to established browsers like **Chrome** and **Edge** (which now feature

By Honghao Wang