<h2>Must know</h2>
<ul>
<li><strong>Google embedding model deprecation causes vendor lock-in</strong>: Google deprecated its <code>text-embedding-004</code> model, rendering thousands of stored vectors unusable for a production RAG-based movie recommender. This highlights a critical issue: different embedding models produce incompatible vector spaces, making migration difficult and creating a hidden form of vendor lock-in. The author, as reported on February 22, 2026, had to re-embed their entire dataset, emphasizing the need to treat embedding models as infrastructure and design for change. <a href="https://app.daily.dev/posts/xhDI1zj4i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel 7.0 rc1 released with succession commentary</strong>: Linus Torvalds released Linux kernel 7.0 rc1, noting the major version bump is purely due to his aversion to large numbers, not a significant feature milestone. The release includes cemented Rust support, faster cache clearing, non-disruptive kernel updates, and updated AMD/Intel silicon support. Torvalds also mused about a future maintainer, referencing the community’s recently developed succession plan, as reported on February 23, 2026. <a href="https://app.daily.dev/posts/cYP8biu57" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Databricks launches Lakebase for AI workloads</strong>: Databricks introduced Lakebase, a new PostgreSQL-compatible database specifically designed for AI workloads. This serverless OLTP database aims to bridge operational capabilities with Databricks’ data lakehouse architecture, separating compute from storage. It supports up to 8TB per instance, includes Postgres 17 with pgvector, and offers features like instant data branching and unified access controls, currently GA on AWS with Azure in public preview, as announced on February 22, 2026. <a href="https://app.daily.dev/posts/56RHAShlB" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Kubernetes 1.35 strengthens AI/ML support</strong>: Kubernetes v1.35, codenamed ‘Timbernetes’, introduces features like workload-aware scheduling with gang scheduling (alpha) and stable in-place Pod resource resize, enhancing its role as an operating system for AI/ML workloads. <a href="https://app.daily.dev/posts/mAyDHBOpw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Rust RwLock can be slower than Mutex for read-heavy workloads</strong>: A hands-on investigation found <code>RwLock</code> performed significantly worse than <code>Mutex</code> for a read-heavy LRU tensor cache in Rust due to cache-line ping-pong from atomic reader counter increments, suggesting <code>Mutex</code> might be better for very short critical sections. <a href="https://app.daily.dev/posts/rFMs5918e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Developer productivity hindered by non-coding tasks</strong>: A developer’s 10-second code fix took over 10 hours due to non-coding tasks like fighting sandboxed test environments, CUDA version conflicts, and build system issues, highlighting that most programmer time is lost to abstraction layers and environmental challenges. <a href="https://app.daily.dev/posts/OBIqfWI7x" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>JDK 24 enhances security with post-quantum cryptography</strong>: JDK 24 introduces significant security enhancements, including post-quantum cryptography (ML-KEM, ML-DSA), improved SHA-3 performance, PKCS#11 AES-CTS support, and the permanent removal of the Security Manager. <a href="https://app.daily.dev/posts/TLN02dr7b" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Shuru: local-first microVM sandbox for AI agents on macOS</strong>: Shuru is a new macOS tool using Apple’s Virtualization.framework to create secure, ephemeral, and reproducible Linux microVMs for AI agent workloads, enabling safe execution of AI-generated code without Docker. <a href="https://app.daily.dev/posts/TqgsBjEoV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Gusto’s universal document processing platform uses AI for abstraction</strong>: Gusto replaced one-off document parsers with a Universal Document Processing (UDP) platform, using AI as a model-agnostic abstraction layer with confidence scores and fallback routing to handle millions of documents annually. <a href="https://app.daily.dev/posts/CDG9JvVKQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Kotlin Multiplatform migration at Masabi</strong>: Masabi, powering public transit ticketing, migrated its decade-old Java codebase to Kotlin Multiplatform, using an incremental approach and agentic IDE tools to unify its tech stack across embedded systems, mobile apps, and backend services. <a href="https://app.daily.dev/posts/leiU4cdzP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s Claude compiler video: more marketing than magic</strong>: Anthropic released a marketing video showcasing Claude supposedly building a C compiler capable of compiling Linux, SQLite, Lua, and Doom. Critics are calling it “dishonest framing,” suggesting the real goal was a stress test of Claude’s autonomous runtime, not a straightforward demonstration of compiler-building prowess. Apparently, we’re still supposed to be impressed by AI that can compile things, even if it’s just for show. <a href="https://app.daily.dev/posts/vGYk0DbTO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Meta patents AI for deceased user simulation</strong>: Meta was granted a patent for an AI system that simulates deceased users’ social media activity, including posting and chatting, trained on their past data. Because nothing says ‘grief counseling’ like your dead grandma’s AI posting about her favorite cat memes. Ethical and mental health concerns? Nah, just another patent for the metaverse. <a href="https://app.daily.dev/posts/AeW0pHCcd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Neiry’s bio-drones: pigeons with neural interfaces</strong>: Neiry, a deeptech company, is testing ‘bio-drones’—live pigeons fitted with neural interfaces, solar panels, and cameras to guide them along preset routes. They claim these birds can cover 400 km daily without battery swaps and are ‘privacy compliant’ by filtering identifiable details. Because what could possibly go wrong with brain-controlled surveillance pigeons? And ‘privacy compliant’ from a company that puts cameras on birds? Sure, Jan. <a href="https://app.daily.dev/posts/m7v8MA0Dh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
