<h2>Must know</h2>
<ul>
<li><strong>Google gemini faces IP theft and nation-state exploitation</strong>: Google’s Gemini AI is facing serious challenges, with over 100,000 prompts attempting to extract its proprietary reasoning capabilities, which Google considers IP theft. Nation-state actors are also integrating Gemini into attack operations for social engineering and phishing. This highlights the dual-use nature of advanced AI and the escalating battle for model security and intellectual property, as reported by Google’s own threat report. <a href="https://app.daily.dev/posts/OQLlwLpSw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 kernel integrates post-quantum crypto and performance boosts</strong>: The Linux 7.0 kernel is shaping up with significant updates, including performance optimizations for memory management and XFS file system, expanded hardware support for Intel Nova Lake and ASUS motherboards, and the integration of ML-DSA, a quantum-resistant signature algorithm. This move towards post-quantum cryptography is a critical step in securing the kernel against future threats, as detailed across multiple technical reports. <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>AI transforms software development and developer roles</strong>: AI tools are fundamentally reshaping software development, eliminating routine coding tasks and shifting the bottleneck to architectural decisions and verification. While senior engineers become more productive “agent bosses,” junior developers face a challenge as entry-level work disappears. This transformation is leading to new roles focused on managing AI assistants and ensuring code quality, as highlighted by reports from Anthropic and Box. <a href="https://app.daily.dev/feed-by-ids?id=0wFgrbFqP&amp;id=78oOrNVDz&amp;id=B3N2Icy7Y&amp;id=BKJ8NHw6d&amp;id=DcSKM9BQA&amp;id=GManH4sq3&amp;id=GnwSkM7qk&amp;id=HHMxTl4yN&amp;id=JgHE2Ywe8&amp;id=MPMZLgvp3" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Minimax launches M2.5 AI model for coding</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 coding tasks. <a href="https://app.daily.dev/posts/VQt2ETdNG" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI releases GPT-5.3-Codex-Spark for faster coding</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. <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>Agentic AI deployment gains traction with elastic and microsoft</strong>: Elastic and Microsoft are pushing agentic AI deployment, with Elastic focusing on data quality for enterprise agents and Microsoft introducing Azure Copilot for continuous cloud optimization. <a href="https://app.daily.dev/feed-by-ids?id=IQCKAyNnH&amp;id=XwFeQESAO" 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 is out, featuring the Green Tea garbage collector by default, enhanced <code>new</code> syntax, 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>New self-hosting tools for paas and AI assistants</strong>: Orchestra, a new open-source self-hosted PaaS, automates server provisioning and application deployment for bare metal, while GoClaw offers a free app for managing self-hosted OpenClaw AI assistants. <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>HBM4 memory enters mass production for AI accelerators</strong>: Samsung and Micron have begun mass production of HBM4 memory, the next-gen high-bandwidth memory for AI accelerators, with Nvidia’s upcoming Vera Rubin accelerators set to utilize it. <a href="https://app.daily.dev/feed-by-ids?id=OUBYQ2PHL&amp;id=WTPyAW1Ca&amp;id=gpkg8awYO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rari a new react framework leverages rust for performance</strong>: A new React framework, rari, launched with Rust under the hood for native-level performance, featuring Rust-powered compilation and built-in React Server Components support. <a href="https://app.daily.dev/feed-by-ids?id=O46mujbgp&amp;id=POxokP60V&amp;id=T7AGr1PoJ&amp;id=WTP9AqFIZ&amp;id=f1ZDIRrLa&amp;id=gPiJ7v5aj&amp;id=jnhJOPqJP&amp;id=r4TGURizp&amp;id=r82S642wz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Gartner discovers AI projects can fail</strong>: Gartner, in its infinite wisdom, predicts 40% of AI projects will fail by 2027 due to “unsustainable velocity” and “shadow AI.” Apparently, they just discovered that throwing AI at every problem without a plan leads to, shocker, failure. Who knew that fragmented systems erode margins and shadow AI creates security vulnerabilities? It’s almost like basic project management and security principles still apply, even with fancy AI. <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>Hackers turn bossware against the bosses (yawn)</strong>: An unnamed threat actor is now exploiting employee monitoring software and remote management tools to deploy ransomware and steal crypto. The article’s title, “Hackers turn bossware against the bosses,” is just a sensationalized way of saying that bad actors are using legitimate tools for nefarious purposes. Because, you know, it’s totally new for attackers to leverage existing software to evade detection. Next, they’ll tell us water is wet. <a href="https://app.daily.dev/posts/ZxyPtzJpI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>MIT to study if AI can help poor people (eventually)</strong>: MIT’s J-PAL, funded by <a href="http://Google.org" target="_blank" rel="noopener nofollow">Google.org</a>, is launching Project AI Evidence to “rigorously evaluate AI solutions for poverty reduction” using randomized controlled trials. Because what the world really needs is more academic studies to prove if AI can actually help people, rather than, say, just deploying solutions. Meanwhile, US immigration agencies are already deploying facial recognition apps without proper privacy assessments, proving that while academics are busy studying, the real world is just doing whatever it wants with AI. <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>
</ul>
