<h2>Must know</h2>
<ul>
<li><strong>DPRK hackers target high profile Node maintainers</strong>: A coordinated social engineering campaign by DPRK-linked actors is targeting maintainers of critical packages like Lodash, Fastify, and Pino to gain repository write access. According to reports from Socket and multiple independent maintainers, attackers use long-term rapport building and fake video calls to deploy malware that bypasses 2FA by exfiltrating session cookies and AWS credentials. This follows a confirmed compromise of Axios, which serves over 100 million weekly downloads, where a malicious dependency was briefly injected into the supply chain. <a href="https://app.daily.dev/feed-by-ids?id=qBJM8nm0c&amp;id=X7PCk5NAZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 kernel halves PostgreSQL performance on ARM</strong>: An AWS engineer discovered that the upcoming Linux 7.0 kernel reduces PostgreSQL throughput by nearly 50% on Graviton4 hardware due to restricted kernel preemption modes. While a patch to restore previous behavior was proposed, kernel maintainers are pushing for PostgreSQL to instead adopt the new Restartable Sequences (RSEQ) time slice extension. As reported by Phoronix, if no middle ground is found before the stable release in two weeks, production database workloads on modern ARM instances face a massive performance regression. <a href="https://app.daily.dev/posts/6AiZJ4eYv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The architectural mismatch of confidential computing</strong>: New research highlights that technologies like Intel SGX and AMD SEV are suffering from a fundamental design flaw where hardware built for single-tenant desktops is being repurposed for multi-tenant clouds. This mismatch has led to predictable side-channel attacks like TDXRay, which can reconstruct LLM prompts from supposedly secure enclaves. Experts argue that while the direction is correct, the current gap between marketing claims of total provider isolation and the reality of physical-access vulnerabilities remains significant. <a href="https://app.daily.dev/posts/G0e9QgsuJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Absurd brings durable execution to pure Postgres</strong>: Armin Ronacher’s new system uses a checkpoint-based replay model in SQL to replace complex deterministic replay engines like Temporal. <a href="https://app.daily.dev/posts/2SBXsxFzn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Karpathy proposes LLM maintained personal wikis</strong>: The pattern uses LLMs to automate the maintenance, linking, and linting of structured markdown files to prevent knowledge rot. <a href="https://app.daily.dev/feed-by-ids?id=7o1BSKg72&amp;id=OhMZS4gFM&amp;id=fEZOib5I2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TurboQuant WASM enables 6x vector compression in browser</strong>: Built with Zig, this library uses relaxed SIMD to compress float vectors to 4.5 bits while preserving dot products for fast similarity searches. <a href="https://app.daily.dev/posts/h7ZO3PZhH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TigerFS mounts PostgreSQL as a standard Unix filesystem</strong>: This FUSE-based tool allows interacting with database rows as files using standard CLI tools like ls and grep, simplifying AI agent integration. <a href="https://app.daily.dev/posts/subDpKF2e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Biome migration cuts linting time by 92 percent</strong>: A case study shows replacing ESLint and Prettier with Biome’s Rust engine reduced pre-commit hooks from 27 seconds to under 2 seconds. <a href="https://app.daily.dev/posts/fsuVXtmDX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TinyOS delivers full network stack in 10KB kernel</strong>: This ultra-lightweight RTOS for ARM and RISC-V includes priority inheritance, MQTT 3.1.1, and TLS 1.3 within a minimal RAM footprint. <a href="https://app.daily.dev/posts/5LG3qkOF4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ryx ORM bypasses Python GIL for database calls</strong>: By using a Rust core via PyO3, this FastAPI-friendly ORM reduced database wait times from 87% to 11% in production testing. <a href="https://app.daily.dev/posts/0TeARnMMJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The AI coding productivity illusion</strong>: While developers feel 20% faster using tools like Cursor, data from DORA and GitClear shows task completion time actually increased by 19% due to a 41% spike in bugs. We’ve traded thoughtful architecture for ‘slopsquatting’—hallucinated dependencies and 1.7x more code churn that downstream teams now have to clean up. It turns out generating code faster than you can review it isn’t a feature; it’s a technical debt accelerator. <a href="https://app.daily.dev/feed-by-ids?id=0NLiScdCr&amp;id=2fJ0hB1Cy&amp;id=o7qcycgq0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Natural language harnesses are the new magic beans</strong>: Tsinghua researchers claim that writing AI agent instructions in plain English outperforms hand-coded Python harnesses by 17%. While the ‘self-evolution’ buzzwords sound great, it’s essentially arguing that removing deterministic guardrails in favor of more LLM vibes is the path to reliability. It’s a bold strategy to solve non-deterministic software problems by adding more non-determinism to the control layer. <a href="https://app.daily.dev/posts/qoeaSTl6M" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
