<h2>Must know</h2>
<ul>
<li>RAM and storage prices skyrocket due to AI demand: Memory and storage prices have surged 2-3x in early 2026, primarily driven by the insatiable demand for high-bandwidth memory (HBM) in AI data centers. DDR4 is phasing out, and HBM production requires three times more wafer capacity than DDR5, straining overall supply. Kingston reports a 246% increase in wafer prices, with forecasts suggesting another 50% rise for DDR5 and SSDs in the first half of 2026, impacting both consumer and enterprise markets. <a href="https://app.daily.dev/posts/RV04f4xGD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Meta secures nuclear energy for AI data centers: Meta has partnered with Vistra, TerraPower, and Oklo to secure up to 6.6 GW of nuclear power by 2035 for its AI infrastructure and data centers. This initiative includes extending the life of existing plants and developing advanced reactors, with a primary focus on Meta’s Prometheus supercluster in Ohio. The move aims to provide clean, reliable baseload power, supporting Meta’s AI expansion and generating thousands of jobs, as announced by Meta. <a href="https://app.daily.dev/posts/elp6Y0W2r" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Amazon’s AI ‘buy for me’ feature sparks merchant outrage: Amazon launched an AI-driven ‘Buy For Me’ feature that scrapes independent merchant websites and lists their products on Amazon without consent. Small business owners reported surprise orders and issues like incorrect product images and public exposure of wholesale pricing. Merchants are criticizing Amazon’s opt-out approach, arguing it unfairly burdens them and damages carefully cultivated customer relationships, especially as Amazon blocks other AI crawlers from its own site. <a href="https://app.daily.dev/posts/HH05G9maY" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li>Python 3.14 promises speed and threading improvements: Python 3.14, slated for 2026, will introduce lazy imports and official free-threading support, potentially offering up to 10x speedups for parallel workloads and addressing long-standing community demands for performance and concurrency. <a href="https://app.daily.dev/posts/quf6PiAcD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>EU open source strategy targets big tech dependence: The European Commission is consulting on an open source strategy to reduce reliance on U.S. tech giants, treating open source as critical infrastructure and shifting funding to scale viable solutions in cloud, AI, and cybersecurity. <a href="https://app.daily.dev/posts/nSTd14JQu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Docker releases free hardened images for security: Docker has made over 1,000 hardened, open-source images available for free under Apache 2.0, aiming to reduce supply chain vulnerabilities by removing unnecessary packages and outdated dependencies from standard Docker base images. <a href="https://app.daily.dev/posts/E2h3E1Ajx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Google launches free AI upskilling platform: Google Skills, a new free online learning platform, offers over 3,000 courses focused on AI and technology, including prompt engineering and machine learning, with personalized paths and Google Career Certificates. <a href="https://app.daily.dev/posts/1LfniMfaa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>New method for encrypted quantum information backup: University of Waterloo researchers developed a method to duplicate encrypted quantum information without violating the no-cloning theorem, enabling secure, redundant quantum cloud storage using expiring one-time encryption keys. <a href="https://app.daily.dev/posts/A8qpCsnhq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Electromagnetic attack on iris recognition systems: Researchers demonstrated EMIRIS, an electromagnetic side-channel attack that extracts biometric iris information from NIR sensors, achieving a 53.47% success rate in spoofing commercial iris recognition systems. <a href="https://app.daily.dev/posts/crEIZ8f1u" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Kilo AI coding agent platform launches: Kilo, an open-source AI coding agent platform, offers integration with major IDEs, access to over 500 AI models, cloud agents, automated code reviews, and a pay-as-you-go model for development teams. <a href="https://app.daily.dev/posts/t9TmWNeD6" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li>AI coding assistants: faster, but worse?: Turns out, those AI coding assistants making developers feel 20% faster actually make them 19% slower and introduce 1.7x more bugs, according to multiple studies. It’s almost like reducing typing isn’t the same as delivering quality code. Organizations are now advised to focus on validation and testing, because apparently, AI amplifies both good and bad practices, and we’re mostly seeing the latter. <a href="https://app.daily.dev/posts/fLWmEtnzQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Poison Fountain: the hero AI deserves, not the one it needs: A shadowy group called Poison Fountain, allegedly from major U.S. tech companies, is encouraging website operators to poison data scraped by AI crawlers. Their goal? To corrupt AI models with subtly incorrect code, citing Anthropic research that a little bit of bad data can cause ‘model collapse.’ Because when you’re worried about existential threats from AI, the obvious solution is to make it dumber, not, you know, regulate it or something useful. <a href="https://app.daily.dev/posts/ULQiNTTAZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Google removes AI overviews from medical search, but only the ones we complained about: Google, in a stunning display of responsiveness, removed AI Overviews from select medical queries after The Guardian pointed out they were giving dangerously misleading health advice. Google claims its internal clinicians found the summaries ‘often accurate,’ which is reassuring, given they only fixed the specific examples called out. So, if you’re asking about something else, good luck with your AI-generated diagnosis! <a href="https://app.daily.dev/posts/5rDA4vp3y" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>