<h2>Must know</h2>
<ul>
<li><strong>Critical RCE in Juniper core routers and AirSnitch WiFi bypass</strong>: Juniper Networks disclosed a critical unauthenticated root RCE (CVE-2026-21902) in PTX series routers, potentially allowing full traffic interception at the core network layer. Simultaneously, UC Riverside researchers introduced AirSnitch, a set of vulnerabilities that bypass WiFi client isolation across vendors like Cisco and Ubiquiti, even on WPA3 networks. These flaws exploit gaps between Layer 2 and Layer 3, enabling machine-in-the-middle attacks despite modern encryption standards. <a href="https://app.daily.dev/feed-by-ids?id=0X8b0bm1J&amp;id=tSnGJTDSb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI hits 730 billion valuation via circular compute deals</strong>: OpenAI secured $110 billion from Amazon, Nvidia, and SoftBank, though much of the capital is tied to massive compute commitments rather than pure cash. Amazon’s $50 billion contribution includes renting two gigawatts of Trainium accelerators, while Nvidia’s $30 billion is linked to deploying five gigawatts of Vera Rubin systems. This structure guarantees returns for the investors’ own cloud and chip businesses while inflating OpenAI’s paper valuation despite profitability not being expected until 2029. <a href="https://app.daily.dev/feed-by-ids?id=8imM4YubL&amp;id=ShShzSJ3B" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google API keys silently granted Gemini authentication</strong>: Truffle Security researchers discovered that GCP API keys, originally intended for billing, were silently upgraded by Google to act as authentication credentials for the Gemini AI API. This change occurred without developer notification, leading to nearly 3,000 exposed keys in the wild that could be used to access private datasets or exhaust quotas. Google initially dismissed the report as intended behavior before acknowledging it as a bug and restricting the affected keys. <a href="https://app.daily.dev/feed-by-ids?id=InhdaYtFz&amp;id=601vxKUqh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Vercel open sources React and Next.js optimization rules for AI agents</strong>: The react-best-practices repository codifies 40+ architectural rules designed for consumption by tools like Cursor and Claude Code. <a href="https://app.daily.dev/posts/J6VGCUvIf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubetorch simplifies remote PyTorch execution on Kubernetes</strong>: A new ecosystem project allows Python ML code to run remotely via a simple .to() API with sub-2-second hot reloading. <a href="https://app.daily.dev/posts/3II62MbL8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft releases Evals for Agent Interop starter kit</strong>: The open-source kit provides a harness to measure schema adherence and tool call correctness for enterprise AI agents. <a href="https://app.daily.dev/posts/0RYNFsgWk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 optimizes slice allocations to reduce GC pressure</strong>: The compiler now uses speculative stack-allocated backing stores for append calls, moving escaping slices to the heap only when necessary. <a href="https://app.daily.dev/posts/9nI5Vn4pu" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NextSilicon Spectra supercomputer enters testing at Sandia Labs</strong>: The system uses 128 reconfigurable Maverick-2 accelerators designed to optimize data flow for HPC workloads that struggle on GPUs. <a href="https://app.daily.dev/posts/t9r2wgx3C" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>India blocks Supabase access under IT Act</strong>: The government issued a blocking order affecting major ISPs, disrupting the developer ecosystem in Supabase’s fourth-largest market. <a href="https://app.daily.dev/posts/x9ggXCarx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kubernetes Ingress NGINX to shut down in March 2026</strong>: CloudNative.Now reports the project is sunsetting, urging users to migrate to the Gateway API immediately. <a href="https://app.daily.dev/posts/oOViN9Ta5" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The 110 billion dollar circular economy</strong>: OpenAI’s latest ‘funding’ is a masterclass in accounting gymnastics where Amazon and Nvidia essentially lend OpenAI money to buy their own products. It’s a closed-loop system designed to pump valuations and guarantee hardware sales while pretending the AI industry isn’t burning cash at an unsustainable rate. Calling this an ‘investment’ is like a landlord giving a tenant a loan specifically to pay the rent. <a href="https://app.daily.dev/feed-by-ids?id=8imM4YubL&amp;id=ShShzSJ3B" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are taking your jobs says company firing 4000 people</strong>: Block (Square) is the latest to blame ‘AI efficiency’ for laying off 10% of its workforce, a convenient narrative that investors gobbled up to the tune of a 24% stock jump. In reality, it’s the same old story of overhiring during the ZIRP era, now rebranded with an AI sticker to make corporate restructuring look like ‘innovation.’ If AI were actually that good at their backend, they wouldn’t need a 14-day discussion window just to change a PHP license. <a href="https://app.daily.dev/feed-by-ids?id=y7FHwZ7mA&amp;id=XIpmjFf08" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The million AI monkeys hypothesis</strong>: The tech hype machine is currently obsessed with the idea that AI agents will soon generate entire production-ready SaaS platforms from a single prompt. This ignores the reality that generating code is the easy part; the real work is the verification, maintenance, and architectural integrity that these ‘monkeys’ consistently fail at. We’re not entering an era of ‘no-code,’ we’re entering an era of ‘infinite technical debt’ where the cost of debugging AI hallucinations will eventually exceed the cost of just hiring a competent engineer. <a href="https://app.daily.dev/feed-by-ids?id=035cEsTZZ&amp;id=trvEPJHeN" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
