<h2>Must know</h2>
<ul>
<li><strong>PHP ecosystem advances with FrankenPHP and Mago</strong>: PHP 8.5 is bringing the pipe operator, clone with parameters, and fatal error backtraces, with pattern matching on the horizon. FrankenPHP, a Go-based runtime, is gaining traction with 10% performance gains and PHP Foundation support. Mago, a Rust-based linter, formatter, and analyzer, has reached a stable release, signaling a continued push for performance and tooling improvements in the PHP ecosystem. <a href="https://app.daily.dev/feed-by-ids?id=lsDO240Ed&amp;id=Z30BC4kwV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is changing software development, not replacing it</strong>: AI is fundamentally transforming software development, shifting the focus from writing code line-by-line to architectural thinking and managing AI agents. While AI can generate significant portions of code and automate tasks, human oversight, domain knowledge, and critical thinking remain crucial. This means developers need to adapt by mastering AI tools, understanding architectural patterns, and focusing on higher-level problem-solving, rather than fearing replacement, as highlighted by InfoWorld and Tech With Tim. <a href="https://app.daily.dev/feed-by-ids?id=1olO8eENO&amp;id=a995yVpWO&amp;id=KP54wqj8l&amp;id=HylUNYAE4&amp;id=fShhVUBWD&amp;id=srWtRgwUR&amp;id=LGCuKx7Wj&amp;id=ra3G8l8Dw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MySQL open-source activity declines, PostgreSQL gains ground</strong>: MySQL’s official GitHub repository has seen no commits since September 2025, following Oracle staff layoffs. This marks a steady decline in open-source activity since 2019, leading industry experts to speculate about Oracle’s shift towards proprietary versions. Meanwhile, PostgreSQL has surpassed MySQL in popularity among professional developers, prompting recommendations to consider alternatives like MariaDB and PostgreSQL, despite the migration challenges. <a href="https://app.daily.dev/feed-by-ids?id=hd9IsUidj&amp;id=UHRxELagh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vinted scales with event-driven architecture</strong>: Vinted’s engineering team successfully transitioned from a monolithic architecture to a global, event-driven platform, handling over 300,000 requests per second by applying Domain-Driven Design and embracing eventual consistency. <a href="https://app.daily.dev/feed-by-ids?id=lUNh7QXHC&amp;id=yEYp1crv1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Linux kernel and QEMU updates streamline infrastructure</strong>: QEMU 11.0 is dropping 32-bit host CPU support, cleaning up 7,300 lines of code, while the Linux kernel 7.0 will remove the legacy mount API, simplifying future development and impacting tools reliant on the old API. <a href="https://app.daily.dev/feed-by-ids?id=50YPqvTFE&amp;id=BqO1pYoWM&amp;id=FlCRi4J5x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents need persistent VMs for secure execution</strong>: <a href="http://Fly.io" target="_blank" rel="noopener nofollow">Fly.io</a> introduced Sprites, durable, stateful Linux VMs designed for secure AI code execution, offering instant checkpointing and persistence, addressing security risks like prompt injection and AI hallucination. <a href="https://app.daily.dev/feed-by-ids?id=Hp5QkrsMN&amp;id=HAEIV3Yfw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered data centers demand new infrastructure</strong>: AI workloads require fundamentally different data center infrastructure, focusing on low-latency GPU-to-GPU communication, 800 Gbps Ethernet, liquid cooling, and power densities of 30-80+ kW per rack, as highlighted by Cisco and MIT Technology Review. <a href="https://app.daily.dev/feed-by-ids?id=cCfgRDo61&amp;id=bf0xB32y7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLVM/Clang adds Ampere CPU support</strong>: The LLVM/Clang open-source compiler now explicitly supports the Ampere1C CPU architecture, enabling developers to optimize code for the upcoming AmpereOne Aurora processor. <a href="https://app.daily.dev/posts/hmTyBIt3G" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Postman acquires Fern for API documentation and SDK generation</strong>: Postman acquired Fern, a startup specializing in API documentation and SDK generation, aiming to reduce developer friction and support costs by providing up-to-date API documentation and SDKs in nine languages. <a href="https://app.daily.dev/posts/9UqlZV9z7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source React calendar and scheduler released</strong>: CalendarKit, a free and open-source React component library, offers month, week, day, agenda, and resource views with features like drag-and-drop scheduling, dark mode, and TypeScript integration. <a href="https://app.daily.dev/feed-by-ids?id=MBDqbsDnk&amp;id=RcOUV5N3R&amp;id=mf2DGEww0" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI-generated code: faster, but at what cost?</strong>: Apparently, AI coding tools are so good they’re causing ‘dead zones’ of 5-30 seconds, forcing engineers to switch contexts and ruin their ‘deep work.’ The solution? An ‘AI Detox’ protocol where you literally remove your hands from the keyboard. Because, you know, the problem isn’t the tools, it’s our inability to stare blankly at a screen for half a minute without checking Slack. Maybe just get faster AI, or, you know, actually read the code it generates? <a href="https://app.daily.dev/feed-by-ids?id=eC9HsHF69&amp;id=infazu71Y&amp;id=nVFL2FYYR&amp;id=uTQFFroPQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Model Context Protocol: the next big thing, or just another fad?</strong>: Anthropic and its pals are pushing this ‘Model Context Protocol’ (MCP) as the holy grail for AI agents to talk to services. Docker and Salesforce are jumping on board, claiming it’ll revolutionize everything. But some folks are calling it a ‘fad,’ saying it adds complexity, security risks, and doesn’t really solve anything that LangChain or a good old script couldn’t. So, is it a genuine leap forward or just another buzzword bingo entry for enterprise AI? <a href="https://app.daily.dev/feed-by-ids?id=6FfwDq5HO&amp;id=7afuAjHXS&amp;id=8827svTqM&amp;id=yBoI44uD0&amp;id=Z30BC4kwV&amp;id=D1vm0YPLg&amp;id=hOkZZEdlQ&amp;id=0u6TE2UfA&amp;id=Nxc0DSNLq&amp;id=zSuC0ti7M" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI leaderboards: optimizing for truthiness, not truth</strong>: LMArena, the AI model leaderboard, is apparently a ‘cancer’ on the industry. Why? Because it relies on casual internet users who prefer ‘confident-looking but incorrect answers’ over actual factual accuracy. So, models are optimizing for superficial qualities like formatting and emojis, pushing the entire AI industry towards ‘appearance over substance.’ Who knew that crowdsourcing truth from the internet would lead to, well, the internet? <a href="https://app.daily.dev/feed-by-ids?id=ViTHGMI9f&amp;id=hTtkIBIga" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
