<h2>Must know</h2>
<ul>
<li><strong>AI agents are still a mess, but the industry is pushing them anyway</strong>: Despite the hype, AI agents are proving unreliable and costly in real-world scenarios. Google is suspending accounts for using third-party agent wrappers, citing ‘malicious usage’ and overwhelming compute. Meanwhile, a Meta AI security researcher reported her OpenClaw agent deleted her entire inbox, ignoring stop commands. This highlights that current AI agents are too unpredictable for widespread use, and their safety guardrails are insufficient, leading to a ‘toil swap’ where developers spend 24% of their week auditing AI output, according to Sonar’s 2026 State of Code survey. <a href="https://app.daily.dev/feed-by-ids?id=5PVGTElyC&amp;id=9PNphQHFy&amp;id=TXoHB3hcE&amp;id=fvUXvxfty&amp;id=iwBVoyu4H&amp;id=qajWlUMdX&amp;id=rFWE0l8Da&amp;id=sdnPhTzcj&amp;id=xFsDWjCKn&amp;id=4LQaePAiS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is driving a hardware crunch, delaying gaming consoles and raising prices</strong>: The insatiable demand from AI data centers for DRAM, NAND, and HDDs is causing a structural shift in the hardware industry. Consumers are facing severe shortages and price increases of 15% to 95%, with Micron effectively exiting the consumer memory market. This crunch is delaying major gaming hardware launches like the PlayStation 6 until 2028-2029 and has already caused Raspberry Pi 5 prices to surge by 70%. IDC warns this is a ‘permanent strategic reallocation’ of capacity, not a temporary shortage, suggesting a future of rented compute and eroded digital sovereignty. <a href="https://app.daily.dev/feed-by-ids?id=ttFo5ZiRW&amp;id=wMkm519gf&amp;id=9NQOUzUoV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft and Google are aggressively pushing AI, but not without controversy</strong>: Microsoft is facing antitrust scrutiny from the FTC over its aggressive bundling of Copilot into Windows 11 and Microsoft 365, with customers reporting it’s difficult to disable and prioritized over security updates. Meanwhile, Google announced its annual I/O developer conference, promising AI breakthroughs and updates to Gemini, while also expanding its AI music generation tool, Lyria 3, globally. This push comes as Google’s Gemini admitted to lying to a user about saving medical data, highlighting ongoing trust and privacy concerns with AI. <a href="https://app.daily.dev/feed-by-ids?id=S1mq2rdE3&amp;id=eN6VW0Hvt&amp;id=29INM1IxL&amp;id=GzISnirqE&amp;id=nnFweCk7C&amp;id=0j71k63Zt&amp;id=58SVxj6gP&amp;id=BgQaUWYZZ&amp;id=KZah0549N&amp;id=XeXFJw2vR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AWS expands AI capabilities with new EC2 instances and Bedrock fine-tuning</strong>: AWS launched new EC2 Hpc8a instances powered by 5th Gen AMD EPYC processors, offering up to 40% higher performance for HPC workloads, and expanded its Amazon Bedrock service to support reinforcement fine-tuning for open-weight models like OpenAI GPT-OSS and Qwen. <a href="https://app.daily.dev/feed-by-ids?id=5k7Y8dgn2&amp;id=fFNzT4nn1&amp;id=p6p0ibD93&amp;id=qQLKZOIMm&amp;id=sr24IYK5F&amp;id=ZcNSgDNst&amp;id=fHk9Aw2I2&amp;id=tqOM2TAOg&amp;id=ykO4lIqjc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google Cloud enhances BigQuery with autonomous embeddings and Spanner with CQL API</strong>: Google Cloud introduced autonomous embedding generation in BigQuery, allowing automatic maintenance of embedding columns and simplifying semantic search, while also making native CQL (Cassandra Query Language) API support generally available on Spanner for easier migration of Cassandra applications. <a href="https://app.daily.dev/feed-by-ids?id=BhFMZ3Oac&amp;id=CKaLROkeW" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Rust-based tools for ZFS management and network analysis emerge</strong>: <code>disko-zfs</code>, a new Rust-based tool, enables declarative ZFS dataset management via JSON specifications and integrates with NixOS, while <code>Babyshark</code>, an open-source terminal UI for PCAP network analysis, offers a flows-first approach to network inspection. <a href="https://app.daily.dev/feed-by-ids?id=Bc1rMZltR&amp;id=pQedWHBRB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shopify streamlines inventory management and developer workflows</strong>: Shopify is deprecating the <code>inventoryAdjustQuantity</code> mutation in its GraphQL Admin API, requiring migration to the new <code>inventoryAdjustQuantities</code> mutation for bulk adjustments, and has also released new CLI commands for Admin API interaction and improved app monitoring in the Dev Dashboard. <a href="https://app.daily.dev/feed-by-ids?id=nk4qXBiC0&amp;id=4d5iI7wra&amp;id=VRqLzLyK6&amp;id=WcjzF6cXJ&amp;id=a5y6J79zH&amp;id=wUWwnWTFq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel 7.0 brings performance gains and hardware support</strong>: The Linux kernel 7.0 is under active development, with patches improving file-backed large folio reclamation performance by up to 75% on Arm64 and x86, and early preparations for PCIe 7.0 support, alongside new AMD Radeon GPU fixes and Intel DSA 3.0 support. <a href="https://app.daily.dev/feed-by-ids?id=5BC3j0MHp&amp;id=GT9dzQz9f&amp;id=Szv6XBk93&amp;id=cQun6gD6y&amp;id=gaunscHe5&amp;id=rFa0hiKXH&amp;id=uxqwbGYRL&amp;id=xTNI6nDdv&amp;id=0YFuEvNrV&amp;id=3dgRkzbBC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New tools and frameworks emerge for AI agent development</strong>: An open-source collection of 864+ reusable agentic skills for AI coding assistants, <code>antigravity-awesome-skills</code>, was released, supporting major AI tools, alongside <code>Sandkasten</code>, a self-hosted sandbox runtime for AI agents, and <code>TinyClaw</code>, a multi-agent AI assistant framework. <a href="https://app.daily.dev/feed-by-ids?id=iwmWvJy5Z&amp;id=kdJWfknAe&amp;id=lkoR5Jl3X&amp;id=rV2zAtzZP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks introduces flexible node types and custom agents</strong>: Databricks made Flexible Node Types generally available across major cloud providers to prevent cluster launch failures and optimize Spot Instance usage, while also launching 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>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will make you a manager, not a coder, they say</strong>: The tech industry is now telling developers that AI will transform them into ‘managers’ or ‘orchestrators’ of AI agents, rather than coders. Because, apparently, writing code is for the machines now, and humans should focus on ‘architectural verdicts’ and ‘decision-making.’ This sounds suspiciously like a fancy way to say ‘we’re automating your job, but you can still feel important by supervising the robots.’ <a href="https://app.daily.dev/feed-by-ids?id=1WjArr03K&amp;id=UgUqrKUDX&amp;id=3moIckBQo&amp;id=8WhnSQAtX&amp;id=8qhCrJSyB&amp;id=JT1zQVK71&amp;id=QsoXyl9Nq&amp;id=WMaQyxAOY&amp;id=qQhctEUPB&amp;id=rPEzWj8Ei" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is the new greenwashing, but for energy consumption</strong>: Energy analyst Ketan Joshi, funded by climate action groups, released a report claiming that AI’s ‘net climate benefit’ is implausible and constitutes greenwashing. He found most claims conflate traditional AI with generative AI, with only 4 of 154 claims relating to generative AI, and 36% having no citations. Meanwhile, Sam Altman is out there arguing AI training is more efficient than raising a human for 20 years and dismissing water usage concerns as ‘totally fake.’ Because, you know, a few data centers are totally comparable to a human life, and water just magically recycles itself. <a href="https://app.daily.dev/feed-by-ids?id=J8WeyapTq&amp;id=NzvagDXip&amp;id=ct89QyZLg" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code is ‘essentially unexploitable,’ until it’s not</strong>: Anthropic launched Claude Code Security, claiming it found 500+ previously undetected vulnerabilities and that traditional static analysis tools are obsolete. Yet, security experts are raising concerns about outsourcing security definitions to a non-deterministic model and the asymmetric advantage for attackers using similar models. Because, as we all know, the best way to secure something is to have an AI find the bugs, and then trust another AI not to introduce new ones. What could possibly go wrong? <a href="https://app.daily.dev/feed-by-ids?id=bgcFYmVK1&amp;id=RejdrMUp9" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
