<h2>Must know</h2>
<ul>
<li><strong>Microsoft’s AI spending spree and Windows 11 woes</strong>: Microsoft’s stock dropped 10% following reports of $37.5 billion in AI-related capital expenditure, raising questions about the ROI of their aggressive AI push. Critics, like those on Hacker News, suggest the focus on embedding Copilot everywhere feels forced, leading to a neglected and bloated Windows 11. This massive investment is happening while the company faces increased emergency patches, potentially linked to workforce reductions and a reliance on AI-generated code, as noted by Windows IT Pro Blog. <a href="https://app.daily.dev/feed-by-ids?id=0PRBbgRZc&amp;id=H1aWwdchu&amp;id=mR53bImT0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Raspberry Pi prices jump due to AI memory demand</strong>: Raspberry Pi announced price increases of $10 to $60 for models with 2GB or more memory, citing doubled LPDDR4 memory costs driven by AI infrastructure demand. This marks a second consecutive year of price adjustments for the same reason, indicating ongoing supply chain pressures in the memory market. Models with 1GB RAM and older LPDDR2 memory are unaffected, according to Raspberry Pi’s official announcement. <a href="https://app.daily.dev/posts/Rk4TtZc2A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Russia-linked APT28 exploits Microsoft Office zero-day</strong>: The Russia-linked APT28 group is actively exploiting a Microsoft Office zero-day vulnerability (CVE-2026-21509) to target Ukrainian government agencies and EU organizations. The attacks use weaponized documents to establish WebDAV connections, download malware, and deploy the COVENANT framework via COM hijacking and scheduled tasks. CERT-UA warns that slow update cycles will likely lead to increased attacks, emphasizing the persistent cyber threat from state-linked actors. <a href="https://app.daily.dev/posts/7E37fbDrv" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI releases open responses specification for agentic AI</strong>: OpenAI released the Open Responses specification to standardize agentic AI workflows, aiming to eliminate API fragmentation and allow developers to switch between models without rewriting integration code. <a href="https://app.daily.dev/posts/yIvolp9mz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub tackles low-quality and AI-generated contributions</strong>: GitHub is addressing the influx of low-quality pull requests, including AI-generated code, by implementing repository-level controls and planning enhanced permission models and AI-powered triage tools. <a href="https://app.daily.dev/posts/jwftmMKjS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent knowledge sharing platforms emerge</strong>: New platforms like Agent Knowledge Commons and Reposit are emerging to enable AI agents to share solutions, learn collectively, and avoid redundant work, offering semantic search and voting on quality. <a href="https://app.daily.dev/feed-by-ids?id=Jq25xqmJD&amp;id=YjbPdZkNb&amp;id=j0iRlHvmi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Firefox 148 introduces granular AI controls</strong>: Firefox 148 will introduce an AI Controls panel, allowing users to disable AI features at a granular level or block them entirely, following user backlash against CPU-intensive and commercially biased AI integrations. <a href="https://app.daily.dev/posts/K3PNUlsQj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux From Scratch drops SysVinit support</strong>: Linux From Scratch (LFS) will become systemd-only starting with version 13.0, citing overwhelming volunteer workload and increasing upstream dependencies on systemd from major desktop environments. <a href="https://app.daily.dev/posts/ohWaYbFxO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust Coreutils progresses towards GNU compatibility</strong>: The uutils project, a Rust implementation of GNU coreutils, released version 0.6 with improved compatibility, reduced unsafe code, and performance enhancements, continuing its goal of 100% GNU compatibility. <a href="https://app.daily.dev/feed-by-ids?id=cuWdYm0D1&amp;id=iBkEhbYAD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Git 2.53.0 released with geometric repacking</strong>: Git 2.53.0 introduces geometric repacking support for partial clone repositories, aiming to maintain separate promisor packfiles and moving closer to becoming the default strategy. <a href="https://app.daily.dev/posts/19pgRer8D" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Cloudflare’s ‘production-grade’ Matrix homeserver</strong>: Cloudflare published a blog post touting a ‘production-grade’ Matrix homeserver built on Workers, which, according to Matrix co-founder Matthew Hodgson, lacked critical features like federation and proper authentication. Turns out, it was mostly unreviewed AI-generated code with TODOs in critical sections. They later added a disclaimer, reclassifying it as a ‘proof of concept.’ Classic overclaiming, folks. <a href="https://app.daily.dev/feed-by-ids?id=46vOfTKm4&amp;id=IgyIzqLcs&amp;id=Iq2RG6GgT&amp;id=bREDMGnKC&amp;id=d6wB2Rm36" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents on Moltbook: 88 bots for every human</strong>: Moltbook, a new social network for AI agents, exposed 1.5 million API keys and private messages due to a misconfigured database. Turns out, 1.5 million registered AI agents were controlled by only 17,000 human users—an 88:1 bot-to-human ratio. And they wonder why there are security issues when you prioritize ‘speed over secure defaults’ in AI-coded applications. <a href="https://app.daily.dev/feed-by-ids?id=iW2dnVT4l&amp;id=nVdzT2QRW&amp;id=IVsgr78CO&amp;id=SulJvcnOu&amp;id=kgAIr1pL1&amp;id=puWFcrOO1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Manis AI: build and launch a mobile app in 8 minutes</strong>: Manis AI claims you can build and launch a mobile app in 8 minutes using their no-code platform and AI orchestration. They demonstrated building a ‘Bill Sniper’ app from a single prompt and deploying it to Google Play with one click. Because, clearly, the hard part of mobile app development was never the design, testing, maintenance, or actual business logic, just the initial drag-and-drop. <a href="https://app.daily.dev/posts/mN5yBPuXJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
