<h2>Must know</h2>
<ul>
<li><strong>Supply chain attacks escalate with AI and CI/CD targets</strong>: A massive surge in software supply chain compromises is hitting the industry, with threat actor TeamPCP exfiltrating 300 GB of credentials from CI/CD pipelines to fuel ransomware operations. Recent attacks on the LiteLLM PyPI package and a critical RCE in the Langflow AI framework (CVE-2026-33017) highlight how quickly vulnerabilities are now being operationalized. Security researchers from Sonatype also warn that AI-generated dependency advice is frequently hallucinated, with GPT-5 inventing nearly 28% of its recommendations, creating a new vector for accidental self-sabotage. <a href="https://app.daily.dev/feed-by-ids?id=DtWPHeTcL&amp;id=HSRYHyvSY&amp;id=KM5ci0LyS&amp;id=PfVtSaby4&amp;id=RgBM7VgFk&amp;id=XVsg2663I&amp;id=cz9svsCYr&amp;id=d7ncWB9vs&amp;id=etL2H8W4G&amp;id=fz5dBqSfO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The death of anonymization in the AI era</strong>: Privacy experts are sounding the alarm as AI makes re-identifying ‘anonymized’ data trivial, leading NYC Health + Hospitals to drop its $4 million contract with Palantir. Research indicates that just six days of step count data can uniquely identify an individual among 100 million people, rendering traditional de-identification methods for health and location data obsolete. This technical shift coincides with federal agencies like the FBI and ICE increasingly bypassing warrants by purchasing bulk personal data directly from commercial brokers, according to reports from privacy advocates. <a href="https://app.daily.dev/feed-by-ids?id=34vaVET2V&amp;id=TkmHu6ESj&amp;id=gwfGGUK0T&amp;id=i0zs41tva" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Intel Xe driver brings 7x performance boost to Linux 7.1</strong>: The upcoming Linux 7.1 kernel will include a major update to the Intel Xe graphics driver that enables Transparent Hugepages (THP) for device pages. This change delivers a 7x speedup for Shared Virtual Memory (SVM) workloads by significantly reducing CPU overhead during page faults. According to Phoronix, the update also introduces new interfaces for querying failed page faults from user-space and optimizations for Xe3P_LPG hardware. <a href="https://app.daily.dev/posts/U9fHRQhXE" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Inertia.js v3.0 drops Axios for a custom XHR client and adds optimistic updates</strong>: The major release slims down bundles by 15KB and introduces a new useHttp hook for reactive, non-navigation requests. <a href="https://app.daily.dev/posts/WWmxp0DEw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Uber migrates to streaming-first data lake with IngestionNext</strong>: The new Flink and Hudi-based architecture reduces ingestion latency from hours to minutes while cutting compute costs by 25%. <a href="https://app.daily.dev/posts/FZIocIP6A" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>RustFS emerges as a viable open-source alternative to MinIO</strong>: Developers are successfully migrating to RustFS on Kubernetes to avoid MinIO’s recent deprecation of open-source features. <a href="https://app.daily.dev/posts/hp0sRpJKU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meilisearch 2026 roadmap targets serverless indexes and hybrid search</strong>: The search engine is evolving into a full retrieval platform with sharding, an AI gateway, and conversational memory for RAG. <a href="https://app.daily.dev/posts/FB8gHvy9S" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub to train AI models on user data by default starting April 24</strong>: The policy change affects Copilot Free and Pro users, including private repositories, though an opt-out mechanism is provided. <a href="https://app.daily.dev/posts/v1oQHNWc2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ATLAS achieves high coding benchmarks on consumer GPUs</strong>: This local inference pipeline uses a 14B quantized model to outperform Claude 4.5 Sonnet on LiveCodeBench at near-zero cost. <a href="https://app.daily.dev/posts/GHvl4XkwS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Systemd developer faces harassment over optional age verification field</strong>: The addition of a self-attested birthDate field to the user database has sparked a toxic debate over FOSS and legal compliance. <a href="https://app.daily.dev/posts/SJccMOCzF" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The intent architect and the collapse of meaning</strong>: Some are now claiming that because AI can vomit out 130k lines of code in two weeks, senior engineers should rebrand as ‘Intent Architects’ who define ‘invariants’ instead of actually building things. It’s a high-brow way of saying you’re going to spend your day writing prompts and hoping the resulting technical debt doesn’t collapse the building. If your ‘architecture’ is just a collection of machine-enforceable OpenAPI validators, you aren’t an architect; you’re a glorified linter manager. <a href="https://app.daily.dev/feed-by-ids?id=C5xlilyRk&amp;id=eTPqMMcbc&amp;id=lgIe84Kc2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Context engineering is just RAG with a better publicist</strong>: The industry is trying to pivot from ‘fragile RAG pipelines’ to ‘Context Engineering’ and ‘Live Data Layers’ as if renaming the problem of stale data makes it go away. We’re being told to ‘pre-compute curated business objects’ for AI agents, which is just a fancy way of describing a materialized view that we’ve been using in databases for decades. It’s the same old data plumbing, now sold with 100% more ‘agentic’ buzzwords to justify the token waste. <a href="https://app.daily.dev/feed-by-ids?id=gIi894ZiM&amp;id=Wn2FSGQOP&amp;id=21ZRiKfwa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered job searching is the ultimate arms race of dishonesty</strong>: We now have developers building custom AI stacks to scrape LinkedIn, score jobs, and ‘dynamically generate’ tailored resumes and cover letters at scale. The author claims this makes them a ‘more self-aware candidate,’ but in reality, it’s just an automated lie-generator designed to bypass the automated lie-detectors used by HR. We’ve reached a point where machines are applying for jobs to be reviewed by other machines, and the humans involved are just there to pay the Vercel bill. <a href="https://app.daily.dev/posts/0YroeqW2V" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
