<h2>Must know</h2>
<ul>
<li><strong>AI agent security risks escalate with new attack vectors</strong>: AI agents are becoming prime targets for sophisticated supply chain attacks, with campaigns like SANDWORM_MODE actively deploying malicious npm packages that target AI coding tools like Claude Code. These attacks harvest credentials and inject malicious code into CI pipelines, even deploying malicious MCP servers to exfiltrate SSH keys. This highlights a critical need for robust security measures beyond traditional code scanning, as AI agents themselves are becoming a significant attack surface, as reported by Socket researchers. <a href="https://app.daily.dev/feed-by-ids?id=3fX3MdYNS&amp;id=tJO4H7DRS&amp;id=OLQboZMRp&amp;id=jkyi3IzXq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft Copilot faces antitrust scrutiny and security flaws</strong>: Microsoft’s aggressive integration of Copilot into Windows 11 and Microsoft 365 is drawing antitrust investigations from the FTC, with concerns about illegal bundling and cloud exit fees. Simultaneously, a confirmed bug in Microsoft 365 Copilot Chat allows it to bypass Data Loss Prevention (DLP) policies and summarize confidential emails, raising serious questions about AI tools circumventing enterprise security controls. This dual pressure highlights the challenges of rapid AI deployment in regulated environments. <a href="https://app.daily.dev/feed-by-ids?id=nnFweCk7C&amp;id=29INM1IxL&amp;id=GzISnirqE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI hardware demand drives memory price surge, impacting consumer tech</strong>: The insatiable demand from AI data centers for DRAM, NAND, and HDDs is causing a structural shift in the hardware industry, leading to severe shortages and price increases of 15% to 95% for consumers. Companies like Micron are exiting the consumer memory market, and major gaming hardware launches like the PlayStation 6 could be delayed until 2028-2029. This isn’t a cyclical shortage but a permanent reallocation of silicon capacity, suggesting a future of rented compute and eroded digital sovereignty for consumers. <a href="https://app.daily.dev/feed-by-ids?id=9NQOUzUoV&amp;id=ttFo5ZiRW&amp;id=wMkm519gf" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Databricks expands cloud offerings with flexible nodes and AI agents</strong>: Databricks made Flexible Node Types generally available across major cloud providers, allowing automatic fallback to alternative instance types to prevent cluster launch failures and optimize Spot Instance usage. They also launched Custom Agents for building and deploying production-quality AI agents on serverless compute. <a href="https://app.daily.dev/feed-by-ids?id=7IXsRxCOl&amp;id=H0KCategS&amp;id=QJn2roYcs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google launches WebMCP for browser-native AI agent tools</strong>: Google introduced WebMCP, a browser-native API enabling AI agents to directly invoke JavaScript tools within Chrome, aiming to reduce token overhead, improve determinism, and narrow the agent attack surface compared to traditional browser automation. <a href="https://app.daily.dev/feed-by-ids?id=61GmU0Tcy&amp;id=DLG34S4jz&amp;id=U2ZYcyKM3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated pull requests overwhelm open source maintainers</strong>: Open source maintainers are increasingly burdened by low-quality, AI-generated pull requests, leading to calls for reputation-based trust mechanisms and specific flagging tools to manage the influx of ‘slop’ content. <a href="https://app.daily.dev/feed-by-ids?id=JAZbbup3r&amp;id=UT5jXEM9J&amp;id=BZXtjLZmd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered SAST tools emerge to combat code vulnerabilities</strong>: Third-generation AI-powered SAST tools combine deterministic rules, dataflow analysis, and LLM-based reasoning to improve vulnerability detection and triage, addressing the high false positive rates of traditional SAST. <a href="https://app.daily.dev/feed-by-ids?id=2TCivC1ph&amp;id=RCpHBctOm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ruby on Rails continues to evolve with AI integration</strong>: The Rails ecosystem is seeing new tools like RubyLLM::Evals for managing LLM prompt configurations and RubyLLM for building AI apps, alongside updates to core Rails features and a new Svelte + Inertia starter kit. <a href="https://app.daily.dev/feed-by-ids?id=4uxvFXtId&amp;id=5tjEzlHaH&amp;id=xllW3fXEF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Datadog shares eBPF hardening lessons from five years at scale</strong>: Datadog engineers shared six lessons from running eBPF-powered workload protection, covering kernel compatibility, safe kernel data capture, consistent eBPF maps, eBPF as an attack surface, tool conflicts, and measuring overhead. <a href="https://app.daily.dev/posts/LuUHFRARO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Rust-based tools enhance Linux and NixOS ecosystems</strong>: New Rust-based tools like <code>disko-zfs</code> for declarative ZFS dataset management in NixOS and <code>Babyshark</code> for terminal-based PCAP network analysis are emerging, showcasing Rust’s growing influence in system-level tooling. <a href="https://app.daily.dev/feed-by-ids?id=DMvnARmzK&amp;id=ZfmQws9TZ&amp;id=v12mxt3cf" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s Claude Code compiler: All hype, no substance?</strong>: Anthropic released a marketing video claiming Claude built a C compiler ‘from scratch’ with ‘no human intervention.’ Turns out, it used 37 years of GCC test suites, an online GCC oracle, and required human restarts. The compiler couldn’t even boot Linux due to a 16-bit x86 code size limitation. It’s almost like AI needs a lot of human help and existing tools to do anything useful, but that doesn’t make for a good press release, does it? <a href="https://app.daily.dev/feed-by-ids?id=0csgEkRMy&amp;id=qulPHX7Sg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI will make you quit programming for a week, apparently</strong>: A developer asked ChatGPT to ‘roast’ their React code and was so ‘terrified’ by the feedback that they ‘quit programming for a week.’ The AI pointed out an 847-line component, 47 <code>useState</code> hooks, and 23 stale TODOs. It’s truly shocking that an AI can identify basic code smells that any human peer reviewer with two brain cells would also spot. Maybe the terrifying part isn’t the AI, but the realization that some developers need an AI to tell them their code is a dumpster fire. <a href="https://app.daily.dev/posts/w3T7LGfKY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>95% of AI pilots fail, but here’s the ‘fix’</strong>: An unnamed author claims 95% of AI pilots fail, based on their ‘experience co-launching 11 AI Proof of Concepts.’ The ‘fix’ is a 12-item checklist covering business alignment, data quality, and human resources. It’s almost as if deploying complex technology without proper planning, coordination, and realistic expectations is a recipe for disaster, regardless of whether it’s AI or a new coffee machine. Who knew? <a href="https://app.daily.dev/feed-by-ids?id=D9Cs0teG0&amp;id=IJOe1UPm1&amp;id=by8MnAaN8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
