<h2>Must know</h2>
<ul>
<li><strong>AI coding agents introduce new 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, accidental 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 Tech Lead Digest. <a href="https://app.daily.dev/posts/3wv1aKzIr" 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 (versions 570.133.20 and 580 series) is causing indefinite system hangs on B200 GPUs running Linux kernel 6.6.0 after approximately 66 days of uptime. The issue, acknowledged by NVIDIA, is linked to NVLink communication failures impacting cross-GPU operations and is suspected to be a jiffies-based timer wrap-around. This bug requires system reboots to restore functionality and currently has no official workaround, impacting operational reliability for those using B200 GPUs. <a href="https://app.daily.dev/posts/xhTAzdzCR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>European Union launches unified company structure and tightens tech regulation</strong>: The European Commission has introduced “EU Inc,” a new company structure allowing startups to operate seamlessly across all EU member states with a single legal entity, aiming to reduce fragmentation and streamline expansion. Simultaneously, the EU is proposing mandatory rules to phase out high-risk technology suppliers from critical infrastructure, reducing dependence on non-EU tech giants like US cloud providers. These moves are part of a broader strategy to enhance digital sovereignty and secure European data, as reported by multiple sources. <a href="https://app.daily.dev/feed-by-ids?id=sCcT14y3N&amp;id=BxeJ5ByjW" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>PostgreSQL adopted as dead letter queue at Wayfair</strong>: Wayfair is using PostgreSQL as a Dead Letter Queue (DLQ) for its event-driven systems, redirecting failed Kafka events to a PostgreSQL table for improved visibility, queryability, and operational simplicity in managing retries. <a href="https://app.daily.dev/posts/eKm9n1imv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google releases jsonschema-go for structured LLM interactions</strong>: Google has launched jsonschema-go, a Go package for JSON Schema that enables schema creation, validation, and inference from Go types, primarily targeting structured interactions with large language models. <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% reduction in runtime, significantly shortening CI pipelines, though memory usage has increased. <a href="https://app.daily.dev/posts/hEVb27hAV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 adds support for Qualcomm Snapdragon 8 Elite Gen 5 and Adreno 225 GPUs</strong>: The upcoming Linux 7.0 kernel will include display support for Qualcomm Snapdragon 8 Elite Gen 5 and Adreno 225 GPUs, thanks to updates in the MSM DRM kernel driver. <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 kernel</strong>: AMD has submitted a series of graphics driver fixes for the AMDGPU and AMDKFD drivers, targeting multiple hardware generations including RDNA4, for inclusion in the Linux 7.0 kernel merge window. <a href="https://app.daily.dev/posts/cxhDtp9rh" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Turbopuffer ANN v3 achieves 200ms p99 latency over 100 billion vectors</strong>: Turbopuffer’s ANN v3 system now boasts 200ms p99 query latency for 100 billion vectors at 1,000 queries per second, 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>FOSDEM 2026 to focus on AI, funding, and geopolitics in open source</strong>: FOSDEM 2026, expecting over 10,000 attendees, will heavily feature discussions on AI’s impact on open source, funding sustainability, and Europe’s digital sovereignty, alongside technical talks on graphics development and the Fediverse. <a href="https://app.daily.dev/posts/TL2tOfQ6b" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding agents are limited by creativity, not technology</strong>: An opinion piece suggests that the main constraint on AI coding agents isn’t their technical capability, but rather programmers’ lack of imagination for innovative uses. Apparently, developers just aren’t creative enough to fully leverage these tools, reducing their impact. So, it’s not the AI that’s the problem, it’s the squishy human brains that can’t keep up. Good to know the blame is squarely on us, not the tech that still struggles with basic context. <a href="https://app.daily.dev/posts/3vC5OgaHL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will end human code writing, say Node.js and Linux creators</strong>: Ryan Dahl (Node.js) and Linus Torvalds (Linux) are apparently in agreement: the era of humans writing code is over. They claim AI will take over direct code generation, shifting human roles to architecture and QA. This sounds suspiciously like the same old ‘X will replace Y’ hype we’ve heard for decades, conveniently ignoring the fact that someone still needs to build, maintain, and debug the AI, and, you know, actually <em>design</em> the systems it’s supposed to be writing code for. Maybe they just want to retire. <a href="https://app.daily.dev/posts/lMnVwtEMR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Personal AI maturity model outlines 9 levels of capability</strong>: Someone has cooked up a 9-level ‘Personal AI Maturity Model’ (PAIMM), categorizing AI from simple chatbots to ‘trusted companion’ assistants that proactively manage your life and career. The top tier, AS3, will apparently monitor your environment and advance your goals without direct prompting. Because what everyone needs is an AI that’s constantly judging your life choices and ‘proactively’ optimizing your existence. What could possibly go wrong with a system that has ‘deep personal understanding’ and environmental control functions? <a href="https://app.daily.dev/posts/Dn9Tp9eLe" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
