<h2>Must know</h2>
<ul>
<li><strong>AI-generated code is still mostly garbage</strong>: Turns out, AI-generated code is often error-prone, insecure, and inefficient, forcing developers to spend more time fixing it than if they’d just written it themselves. Startups are even hiring developers specifically to clean up this ‘AI slop,’ according to various reports. It seems training on average-quality public repos makes AI proficient at boilerplate, but not much else, leading to increased technical debt.</li>
<li><strong>Netlify adds AI agents and multi-model gateway</strong>: Netlify is rolling out Agent Runners, AI agents designed to automate code fixes, updates, and deployments directly from their dashboard, supporting models like Claude, OpenAI, and Gemini. They also launched an AI Gateway to consolidate inference management across various AI provider APIs. This aims to streamline developer workflows, though details on cost and usage limitations are conspicuously absent.</li>
<li><strong>Security flaws plague AI systems and apps</strong>: Multiple security incidents highlight ongoing vulnerabilities, from an event planning app exposing user GPS data to AI browsers being vulnerable to ‘CometJacking’ prompt injection attacks. Even OpenAI’s Sora reportedly has a jailbreak vulnerability exposing sensitive audio. These incidents, covered by TechCrunch and others, underscore the rushed deployment of AI without robust security testing.</li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>C# readonly vs const: a performance gotcha</strong>: A production incident revealed that improper use of <code>readonly</code> in C# can cause unexpected memory allocations and performance degradation, a subtle but critical difference from <code>const</code> that many developers overlook.</li>
<li><strong>Functional programming’s impact on frontend development</strong>: The embrace of functional paradigms fundamentally altered frontend development, leading to practices like CSS-in-JS and synthetic events that sometimes fight native browser standards, though newer tools like HTMX and Astro are shifting back to web standards.</li>
<li><strong>Ada and Rust: different paths to safety</strong>: A comparison of Ada and Rust in solving Advent of Code puzzles highlights their distinct approaches to safety: Ada emphasizes compile-time correctness and high-level types, while Rust focuses on memory safety via ownership and Result/Option types.</li>
<li><strong>Intel orphans Linux software packages</strong>: Intel has orphaned numerous software packages on Debian and Ubuntu, including tools for accelerators, performance monitoring, and power management, following internal restructuring, leaving their future support uncertain.</li>
<li><strong>Linux 6.18 brings persistent caching and IOMMU updates</strong>: The Linux 6.18 kernel merges new DeviceMapper changes, including DM-PCACHE for persistent block device caching, and IOMMU driver updates for Intel, AMD, Apple, and RISC-V, enhancing virtualization and memory management.</li>
<li><strong>Piwik Pro ends free plan, impacting open-source tools</strong>: Piwik Pro eliminated its free plan, introducing a minimum annual charge of €420, directly impacting individual and open-source developers who relied on no-cost access for testing and support.</li>
<li><strong>India’s tech talent pipeline is sputtering</strong>: India’s entry-level tech hiring has plummeted, with AI automation eliminating routine roles and Global Capability Centers preferring experienced specialists, forcing universities to overhaul curricula towards AI/ML skills.</li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Hard work alone won’t get you promoted, apparently</strong>: According to <a href="http://swizec.com" target="_blank" rel="noopener nofollow">swizec.com</a>, just grinding away isn’t enough for career advancement; you need to identify and fix ‘systemic gaps’ and build tools that ‘multiply team effectiveness.’ Because, you know, doing your job well is for suckers. The real trick is to make sure management notices your ‘critical work’ and aligns it with ‘organizational goals,’ which sounds suspiciously like more work, but with extra steps and buzzwords.</li>
<li><strong>Flock’s gunshot detectors now listening for human screams</strong>: Flock Safety’s Raven system, originally for gunshot detection, is now expanding to identify human distress sounds like screaming, raising significant privacy concerns and legal questions about audio surveillance. Because what could possibly go wrong with AI-powered microphones listening to everything in public spaces, especially when the system already has a 99% false alarm rate? I’m sure this will only lead to more ‘safety’ and definitely not more police shooting at kids mistaking fireworks for gunfire, as happened in Chicago.</li>
<li><strong>Senior engineers: the new corporate puppeteers</strong>: Hacker News suggests that staff software engineers can ‘influence tech company politics’ by aligning their technical projects with ‘organizational priorities’ and timing their proposals to ‘executive attention shifts.’ So, instead of just building cool stuff, you’re now expected to be a corporate strategist, maintaining a ‘portfolio of ready-to-execute technical improvements’ to subtly manipulate the C-suite. Because nothing says ‘engineering’ like playing 4D chess with your boss’s boss.</li>
</ul>
