<h2>Must know</h2>
<ul>
<li><strong>Google Gemini faces IP theft and nation-state exploitation</strong>: Google blocked over 100,000 attempts to extract proprietary reasoning capabilities from its Gemini AI model, which it considers intellectual property theft. Nation-state actors are also integrating Gemini into attack operations for social engineering and phishing. This highlights the dual challenge of protecting advanced AI models from IP theft and preventing their misuse in cyberattacks, as detailed in Google’s threat report. <a href="https://app.daily.dev/posts/OQLlwLpSw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 integrates post-quantum cryptography</strong>: The Linux 7.0 kernel is set to integrate ML-DSA, a quantum-resistant signature algorithm, for module signing, replacing SHA-1. This move is part of a broader industry shift towards post-quantum cryptography, driven by NIST standards, to protect against future quantum attacks that threaten current RSA and ECDSA algorithms. This is a significant step in securing critical infrastructure against emerging threats. <a href="https://app.daily.dev/feed-by-ids?id=6DJe6t3uU&amp;id=Ukr75F53X&amp;id=Vay0ocYOF&amp;id=Vh5TboDqs&amp;id=m5L4rWAXQ&amp;id=ppwBwOuN5&amp;id=uGdXnDZH4&amp;id=xVcPYGmKq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Odido data breach impacts 6.2 million customers</strong>: Dutch mobile operator Odido disclosed a data breach affecting 6.2 million customers, roughly one-third of the Netherlands’ population. Attackers accessed their customer contact system, stealing names, addresses, phone numbers, dates of birth, bank account numbers, and ID details. This incident underscores the ongoing vulnerability of large telecom providers to significant data breaches, as reported by Odido. <a href="https://app.daily.dev/posts/3j2lBB9LH" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>OpenAI launches faster coding model on Cerebras hardware</strong>: OpenAI released GPT-5.3-Codex-Spark, a new coding model running on Cerebras CS3 accelerators, prioritizing speed with 15 times faster inference than previous models on Nvidia infrastructure. <a href="https://app.daily.dev/feed-by-ids?id=2DKu3o06c&amp;id=2G5oHg0oJ&amp;id=A9s0KznyU&amp;id=CacFMCP2L&amp;id=PFJ429xnc&amp;id=TgVTzT7ZG&amp;id=dqsxZt8l5&amp;id=f8bwmVFOm&amp;id=iVwXrkdSj&amp;id=yYXkpJGbb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Minimax launches M2.5 AI model at competitive price</strong>: Minimax launched its M2.5 frontier AI model, claiming it’s over 90% cheaper than Claude Opus or GPT-5 models and excels at software engineering tasks, with OpenHands benchmarking it as a top open model for coding. <a href="https://app.daily.dev/posts/VQt2ETdNG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>HBM4 memory production ramps up for AI accelerators</strong>: Samsung and Micron have begun mass production of HBM4 memory, the next-generation high-bandwidth memory for AI accelerators, with Micron’s 2026 output already pre-sold, indicating high demand and potential price increases for consumer memory. <a href="https://app.daily.dev/feed-by-ids?id=OUBYQ2PHL&amp;id=gpkg8awYO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 released with Green Tea GC and performance boosts</strong>: Go 1.26 was released, featuring the Green Tea garbage collector enabled by default, a new <code>new</code> syntax for initial values, and a 30% reduction in <code>cgo</code> call overhead. <a href="https://app.daily.dev/feed-by-ids?id=1ObNpwyIZ&amp;id=mj2oUwkea" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Orchestra offers self-hosted PaaS for bare metal</strong>: Orchestra, an open-source, self-hosted PaaS, was released to automate server provisioning, cluster orchestration, and application deployment for bare metal infrastructure without requiring agents. <a href="https://app.daily.dev/feed-by-ids?id=0nUdXf1b9&amp;id=C2t5MyKnp&amp;id=jNBhBoH4V&amp;id=nwGZvk6OS&amp;id=sjdf7M4Ia" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare open-sources Rust library for zero-downtime upgrades</strong>: Cloudflare open-sourced ecdysis, a Rust library enabling zero-downtime service upgrades through graceful process restarts, which has been in production for five years handling billions of requests. <a href="https://app.daily.dev/posts/tdBY3TVs4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rails updates include transaction customization and performance fixes</strong>: Rails received updates including a new extension point for customizing transactions, performance improvements for <code>filter_parameters</code>, and fixes for Marshal deserialization and composite primary key handling. <a href="https://app.daily.dev/feed-by-ids?id=5hfEVLtRT&amp;id=CWjTT1lwV&amp;id=oKSotbpAp" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Gartner predicts 40% AI project failure rate</strong>: Gartner predicts 40% of AI projects will fail by 2027 due to ‘unsustainable velocity, fragmented systems, and shadow AI.’ Apparently, throwing AI at every problem without a coherent strategy or proper governance isn’t a magic bullet. Who knew that complex technology requires careful planning and integration? Shocking, I tell you. <a href="https://app.daily.dev/feed-by-ids?id=NUfU1KApc&amp;id=Ss5yJGdsu&amp;id=ZY9MBJa3v&amp;id=myl2tyVQW&amp;id=sK6lOaq5D" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>US immigration agencies deploy flawed facial recognition without oversight</strong>: ICE and CBP deployed NEC’s Mobile Fortify facial recognition app without required Privacy Impact Assessments, even though they knew it couldn’t verify identities as claimed. They then used it to scan US citizens, protesters, and observers. Because nothing says ‘trust us with your data’ like bypassing privacy reviews and deploying tools that don’t even work as advertised, all while scanning everyone in sight. <a href="https://app.daily.dev/feed-by-ids?id=ab5QuRrVe&amp;id=cxpq72Re5&amp;id=lLSTCWQDZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI hype leads to disappointment and distrust</strong>: AI companies like OpenAI, Nvidia, and Anthropic are being accused of setting ‘impossibly high expectations’ with their hyperbolic claims, leading to widespread disappointment and declining public trust. Corporate AI adoption also shows a 95% failure rate in generating measurable value. It’s almost as if promising the moon and delivering a rock isn’t a sustainable business model. But hey, at least the valuations are still astronomical, right? <a href="https://app.daily.dev/feed-by-ids?id=1ssjvObXV&amp;id=6C3DC1GxE&amp;id=7qf6CcnTl&amp;id=KBiI9wOrx&amp;id=NO2cmpN4i&amp;id=PTlBRJtuz&amp;id=PryOnuBPQ&amp;id=Sfm1OU6it&amp;id=ZJJH0MAGc&amp;id=mcpWOltJw" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
