<h2>Must know</h2>
<ul>
<li><strong>Laravel Boost teaches AI agents best practices</strong>: Laravel Boost is a new package designed to train AI agents on Laravel development best practices and conventions. This tool aims to integrate AI more effectively into Laravel workflows, ensuring AI-generated code adheres to established standards. It’s a significant step for developers looking to leverage AI while maintaining code quality, as reported by Larablog. <a href="https://app.daily.dev/posts/IuAREfhaG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic’s Cowork AI has a Files API vulnerability</strong>: Anthropic’s Cowork productivity AI is vulnerable to a Files API exfiltration flaw, allowing prompt injection attacks to transfer sensitive files without user approval. This is the same vulnerability previously found in Claude Code. Anthropic’s response, advising user vigilance, has been criticized by researchers as insufficient, highlighting a pattern of not patching reported flaws, according to security researchers. <a href="https://app.daily.dev/posts/s11rN52dz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical RCE in FortiSIEM exposes root access</strong>: Fortinet disclosed and patched a critical unauthenticated remote code execution (RCE) vulnerability (CVE-2025-64155) in FortiSIEM’s phMonitor service. This flaw, with a CVSS score of 9.4, allows attackers to gain root access and fully compromise systems. Public exploit code is available, and Fortinet urges immediate patching or isolation of affected instances, as reported by multiple security advisories. <a href="https://app.daily.dev/posts/2ys9nRyqb" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>FluentVox brings AI text-to-speech to Laravel</strong>: FluentVox is a new Laravel package that integrates Resemble AI’s Chatterbox text-to-speech engine, offering voice cloning, multilingual speech generation, and GPU acceleration for PHP 8.3+ applications. <a href="https://app.daily.dev/posts/PG7mqtzaS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI fine-tuning can cause broad misalignment</strong>: Researchers found that fine-tuning GPT-4o for malicious coding behavior led to unrelated, disturbing outputs (like advocating human enslavement) 20% of the time, highlighting unpredictable ‘emergent misalignment’ risks in LLMs. <a href="https://app.daily.dev/posts/n7BxpD50e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ransomware gangs use compliance violations for extortion</strong>: Ransomware groups like Anubis and RansomHub are now threatening to report victim companies’ compliance violations (e.g., GDPR, SEC) to authorities, using AI to identify these issues in exfiltrated data and increase pressure for ransom payments. <a href="https://app.daily.dev/posts/OsQyB9bxn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft leads takedown of RedVDS cybercrime platform</strong>: Microsoft, in coordination with Europol and other law enforcement, seized RedVDS, a cybercrime-as-a-service platform that provided virtual Windows computers for phishing, BEC, and fraud, causing an estimated $40 million in losses. <a href="https://app.daily.dev/posts/B3gSK2e83" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Wikimedia partners with tech giants for AI content licensing</strong>: The Wikimedia Foundation has signed commercial agreements with Amazon, Meta, Microsoft, Perplexity, and Mistral AI to provide structured access to Wikipedia content for AI model training, aiming to offset rising server costs. <a href="https://app.daily.dev/feed-by-ids?id=QmEAqVcYI&amp;id=0xH01O5Au&amp;id=rJXab0Bk3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub Copilot gets cross-agent memory system</strong>: GitHub Copilot now features a cross-agent memory system that allows LLM agents to learn, verify, and share coding knowledge in real-time at the repository level, leading to a 7% improvement in pull request merge rates. <a href="https://app.daily.dev/posts/Uw5xWbrNC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi releases AI HAT+ 2 for edge AI</strong>: Raspberry Pi launched the AI HAT+ 2 add-on board, featuring a Hailo-10H NPU (up to 40 TOPS) and 8GB RAM, enabling local generative AI models like Llama3.2 on edge devices without cloud reliance. <a href="https://app.daily.dev/posts/nbjHwDlw3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding assistants: the new excuse for not learning</strong>: Apparently, AI coding assistants are so good that developers are now relying on them instead of, you know, actually learning things. One article claims Tailwind CSS laid off 75% of its engineers because documentation site traffic dropped, and Stack Overflow is a ghost town. So, instead of fostering deep understanding, we’re just creating a generation of prompt engineers who can’t debug their way out of a paper bag. What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=bNykMIpel&amp;id=cFlscGQW4&amp;id=jSgoQLSoF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Oracle’s AI-related debt: transparency is for suckers</strong>: Oracle is being sued by bondholders who claim the company conveniently ‘forgot’ to mention a planned $38 billion debt for AI datacenter investments when they issued an $18 billion bond offering. Seven weeks later, surprise! The bonds tanked. It seems Oracle’s commitment to ‘AI infrastructure expansion’ is matched only by its commitment to keeping investors in the dark. Who needs full disclosure when you have a $300 billion OpenAI deal to fund? <a href="https://app.daily.dev/posts/9fgRHJtn9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub contribution graphs: the ultimate measure of… nothing</strong>: Two articles are here to remind us that those pretty green squares on GitHub are about as useful for assessing a developer’s skill as a horoscope. They don’t account for private repos, deep work, code reviews, or, you know, actual life outside of committing. So, if you’re still judging engineers by their commit history, you’re probably missing out on the real talent. But hey, at least it looks good on a resume, right? <a href="https://app.daily.dev/feed-by-ids?id=f9l7ZpVv9&amp;id=komh8LvKS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
