<h2>Must know</h2>
<ul>
<li><strong>AI coding assistants demand more upfront planning</strong>: A 2025 study found developers using AI tools took 19% longer to complete tasks, despite feeling faster. This is because AI assistants lack contextual memory, leading to inefficient iterations without detailed specifications. Code review times have increased by 91% with high AI tool use, and 45% of AI-generated code contains security vulnerabilities, according to an unnamed study. This means effective AI use requires treating it like a junior developer: lots of oversight and explicit guidance. <a href="https://app.daily.dev/posts/pYjmmMIkn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Netflix mobile app redesign and acquisition ambitions</strong>: Netflix is overhauling its mobile app in late 2026 to feature short-form vertical videos, similar to TikTok, and expanding into live sports with WWE, NFL, and MLB. The company is also reportedly pursuing an $82.7 billion acquisition of Warner Bros Discovery to boost its content library and theatrical distribution, as announced by Netflix. <a href="https://app.daily.dev/posts/4L22WBSq0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>LLM inference hardware bottlenecks shift to memory</strong>: LLM inference is hitting a wall, but it’s not the GPUs. A preprint from January 25, 2026, highlights that the autoregressive Decode phase is bottlenecked by memory capacity and bandwidth, not compute. Future datacenter AI accelerators will need high-bandwidth flash, processing-near-memory architectures, 3D memory-logic stacking, and low-latency interconnects to keep up with ever-growing models. <a href="https://app.daily.dev/posts/t0LNVAkom" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Open-source tools for local AI agent deployment</strong>: A new article lists 20 free and open-source tools like Ollama, LangGraph, and LlamaIndex that enable local deployment of AI agents, offering production-grade capabilities without relying on paid LLM APIs. <a href="https://app.daily.dev/posts/gTbPMcu2k" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Turbopuffer ANN v3 achieves low-latency vector search</strong>: Turbopuffer’s ANN v3 now delivers 200ms p99 query latency at 1k QPS over 100 billion vectors, using hierarchical clustering, binary quantization, and distributing data across storage-dense VMs. <a href="https://app.daily.dev/posts/EHAeemnne" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI microservices architecture for LLM applications</strong>: LLM applications are moving to microservices, using agentic orchestration with tools like LangGraph as a centralized state machine and semantic routing to enhance intent recognition, improving scalability and fault tolerance. <a href="https://app.daily.dev/posts/eRP6IF0kI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Decentralized and retro internet alternatives gain traction</strong>: The IndieWeb and Small Web movements, alongside a revival of Bulletin Board Systems (BBS), are gaining attention as users seek more control and autonomy away from centralized web platforms. <a href="https://app.daily.dev/feed-by-ids?id=EWw76Iimy&amp;id=pYFIL8byT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OnePlus firmware update deploys irreversible anti-rollback</strong>: OnePlus has pushed a firmware update for its 13, 15, and Ace 5 series that blows Qfprom electronic fuses, making it impossible to roll back to older firmware or install custom ROMs without hard-bricking the device. <a href="https://app.daily.dev/posts/gVLCy88Eu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google announces high-friction sideloading for Android</strong>: Google is implementing a ‘high-friction’ sideloading process on Android, adding more steps for users to install apps from unverified developers, framing it as an ‘Accountability Layer’ to increase risk awareness. <a href="https://app.daily.dev/posts/ICjHMVGeX" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>FOSDEM 2026 to focus on AI, funding, and politics in open source</strong>: FOSDEM 2026, expecting over 10,000 attendees, will feature discussions on AI’s impact on open source, funding sustainability, digital sovereignty in Europe, and open source contributor burnout. <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-enabled fraud is a rapidly growing threat, apparently</strong>: AI-enabled fraud is now a ‘rapidly growing threat’ in cybersecurity, with over $300 billion lost to pandemic-relief fraud and $12.5 billion in consumer fraud in 2024. The article, dated January 26, 2026, urges security leaders to shift focus to fraud prevention, because apparently, AI just makes everything more expensive and complicated, even crime. Who knew? <a href="https://app.daily.dev/posts/9mIDnPhWj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Agentic AI automates compliance, because humans are too slow for regulations</strong>: Agentic AI is being integrated into financial and ESG reporting to handle ‘increasingly complex regulatory requirements’ under frameworks like CSRD and ISSB. The articles, all published on January 26, 2026, consistently frame this as transforming compliance from a cost center to a strategic advantage. Because nothing says ‘strategic’ like automating the paperwork nobody wants to do, without a single mention of actual deployment or real-world effectiveness. <a href="https://app.daily.dev/feed-by-ids?id=Skhkgwj1Z&amp;id=qF5kIahRI&amp;id=uTj3yEY4b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ChatGPT and Claude are citing ‘Grokipedia,’ because why not?</strong>: ChatGPT’s GPT-5.2 and Anthropic’s Claude are reportedly citing ‘Grokipedia,’ an AI-generated encyclopedia by Elon Musk, as a source for obscure topics. Grokipedia has faced criticism for ‘inaccurate and biased content,’ but hey, if it’s AI-generated, it must be good, right? This raises questions about vetting sources, but I’m sure the models are just trying to be ‘broadly informed.’ <a href="https://app.daily.dev/posts/SDpas3mop" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
