<h2>Must know</h2>
<ul>
<li><strong>AI coding agents introduce security and governance risks</strong>: The widespread adoption of AI coding agents, now embedded in chat platforms like Slack, is democratizing code creation but also exposing organizations to significant security threats. With 85% of developers using AI tools and 41% of code being AI-generated, risks like prompt injection, credential exposure, and unreviewed code deployments are rampant. Organizations need strict governance frameworks, tiered permissions, mandatory human approval for critical changes, and comprehensive audit trails to manage these new risks, as highlighted by Security Boulevard. <a href="https://app.daily.dev/posts/3wv1aKzIr" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>iOS AI apps are leaking sensitive user data</strong>: Security researchers have identified 198 iOS apps leaking sensitive user data, with AI-related apps being the worst offenders. One app, ‘Chat &amp; Ask AI,’ exposed 380 million private messages, impacting 18 million users by leaking phone numbers and email addresses. CovertLabs documented these flaws, raising serious questions about Apple’s App Store vetting process and the hurried development cycles of AI apps. This highlights a critical security gap in the rapidly expanding AI application landscape. <a href="https://app.daily.dev/posts/d7LHQS5Gm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA B200 GPU driver bug causes system hangs</strong>: NVIDIA’s open GPU kernel driver (version 570.133.20) is causing indefinite system hangs on B200 GPUs with Linux kernel 6.6.0 after approximately 66 days of uptime. This issue, traced to NVLink communication failures, affects cross-GPU operations and has been consistently reproduced in production environments. NVIDIA has acknowledged the bug and is working on a fix, but no workaround is available yet, forcing reboots to restore functionality. <a href="https://app.daily.dev/posts/xhTAzdzCR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Google releases JSON Schema package for Go</strong>: Google has launched <code>jsonschema-go</code>, a new Go package for JSON Schema creation, validation, and inference, primarily aimed at defining structured interactions with large language models (LLMs). <a href="https://app.daily.dev/posts/QUtLwKAlO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PHPStan 2.1.34 delivers major performance upgrades</strong>: PHPStan 2.1.34, a static analysis tool for PHP, has been released with 25-40% faster analysis times on real-world projects and up to 50% runtime reduction, significantly shortening CI pipelines. <a href="https://app.daily.dev/posts/hEVb27hAV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Symfony releases January 2026 patch updates</strong>: Symfony published patch releases (6.4.32, 7.3.10, 8.0.4) with numerous bug fixes addressing issues like memory leaks, dependency injection, and HTTP cache, enhancing stability and backward compatibility. <a href="https://app.daily.dev/posts/4utDSg8Uh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Turbopuffer ANN v3 achieves low latency on massive datasets</strong>: Turbopuffer’s ANN v3 system now boasts 200ms p99 query latency for 100 billion vectors at 1,000 QPS, utilizing hierarchical clustering and 16-32x compression with a 92% recall rate. <a href="https://app.daily.dev/posts/EHAeemnne" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 adds Qualcomm Snapdragon 8 Elite Gen 5 support</strong>: The upcoming Linux 7.0 kernel will include support for Qualcomm Snapdragon 8 Elite Gen 5 display hardware and the Adreno 225 GPU, enhancing graphics capabilities for Linux systems on Qualcomm chipsets. <a href="https://app.daily.dev/posts/LVxtNaVxq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AMD submits graphics driver fixes for Linux 7.0</strong>: AMD has submitted a set of graphics driver fixes for the AMDGPU and AMDKFD drivers, targeting the Linux 7.0 kernel merge window, covering RDNA4 and other hardware generations with improvements to HDMI, VCN, and GPUVM. <a href="https://app.daily.dev/posts/cxhDtp9rh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>SVT-AV1 4.0 delivers significant encoding performance gains</strong>: SVT-AV1 4.0 has been released, offering 5-8x speedup for still image encoding, 5-15% for RTC modes, and 10-25% for Random Access mode, alongside new visual quality enhancements like AC Bias. <a href="https://app.daily.dev/posts/HOUGYgKhJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding agents struggle with novelty, blame the user</strong>: Apparently, AI coding agents aren’t generating novel solutions because programmers ‘lack ideas for innovative uses,’ according to RUBYLAND. So, it’s not the AI’s fault for being a statistical parrot, it’s <em>your</em> lack of creativity. Maybe if we just described checker pieces as ‘UFOs,’ the LLM would finally understand true innovation. Classic blame-the-user strategy. <a href="https://app.daily.dev/posts/gov03HPhm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Encrypted quantum cloning breakthrough, details TBD</strong>: A new quantum protocol claims to enable ‘encrypted cloning of quantum states,’ supposedly circumventing the no-cloning theorem, a foundational limit in quantum mechanics. The article, dated January 25, 2026, touts ‘potential impact’ for quantum data storage and security, but conveniently omits any specific names, institutions, or technical details on how this ‘breakthrough’ actually works. Sounds like someone’s trying to get ahead of the peer review process with a bold claim and zero evidence. <a href="https://app.daily.dev/posts/z7OYuDS7t" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The era of humans writing code is over, says Node.js creator</strong>: Ryan Dahl, creator of Node.js, and even Linus Torvalds, are apparently declaring the ‘era of humans writing code is over,’ with AI taking over direct code generation. This, of course, means developers will now focus on ‘system architecture and code quality assurance.’ So, we’re just going to trust the AI to write the code, and then we’ll… check its work? Sounds like a fantastic way to train junior developers for a future where they just debug AI’s hallucinations. <a href="https://app.daily.dev/posts/lMnVwtEMR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
