<h2>Must know</h2>
<ul>
<li><strong>AWS doubles down on cloud and AI infrastructure</strong>: AWS reported its strongest quarterly growth in over three years, with $35.6 billion in Q4 2025 revenue, a 24% year-over-year increase. Driven by traditional cloud migration and AI workloads, AWS plans to invest $200 billion in 2026 to double its data center capacity by the end of 2027, primarily for AI infrastructure. CEO Andy Jassy stated AWS is monetizing AI capacity as fast as it’s deployed, with custom Trainium accelerators and Graviton CPUs generating $10 billion in annualized revenue. <a href="https://app.daily.dev/posts/1BAlXIloL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are disrupting the SaaS business model</strong>: AI agents are fundamentally shifting the SaaS landscape from selling software tools to delivering outcomes directly. Anthropic’s Cowork plugins, which allow domain expertise to be expressed as simple configuration files, triggered significant stock drops in major SaaS companies. This trend, predicted by Satya Nadella, suggests business logic is migrating to AI tiers, making traditional SaaS applications increasingly obsolete and pushing towards outcome-based pricing models. <a href="https://app.daily.dev/posts/zMRAHcD2R" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft open-sources Rust-based sandboxing project litebox</strong>: Microsoft has open-sourced LiteBox, an experimental library OS written in Rust, designed for application sandboxing with a minimal attack surface. Positioned between containers and VMs, LiteBox offers stronger isolation without heavy overhead by utilizing a library OS model with narrow interfaces. It supports AMD SEV-SNP for confidential computing and can run unmodified Linux applications across different platforms, though it’s currently experimental and not production-ready. <a href="https://app.daily.dev/posts/jIlb9nmzq" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Postgres extensions can replace specialized databases</strong>: For 99% of companies, PostgreSQL with extensions like pg_textsearch, pgvector, and TimescaleDB can handle full-text search, vector search, time-series, caching, and document storage, reducing operational complexity and infrastructure overhead. <a href="https://app.daily.dev/posts/QtM7zNGDf" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools transform development but shift bottlenecks</strong>: AI tools like Cursor and Claude Opus 4.6 are dramatically increasing code generation speed, but overall development velocity is not increasing proportionally. The bottleneck has shifted to coordination, reviews, and wait times between handoffs, highlighting the need for process changes and end-to-end ownership. <a href="https://app.daily.dev/feed-by-ids?id=1UoQrsDRG&amp;id=3IeScMbvD&amp;id=1S4Xqhe7b&amp;id=BJDSL4jOH&amp;id=BQ8poB6iB&amp;id=Wbu0BhyXh&amp;id=aKwpYmoPc&amp;id=bheFZPTb3&amp;id=onAnCwgEm&amp;id=yIPmtTTfY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snowflake integrates pg_lake for open data standards</strong>: Snowflake integrated pg_lake, an open-source PostgreSQL extension, into its Snowflake Postgres service, enabling direct SQL interaction with data lakehouses and Apache Iceberg tables, and expanding its Horizon Catalog for governance across various data formats. <a href="https://app.daily.dev/posts/kBWxo6cQz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apache Flink agents 0.2.0 brings event-driven AI to streaming</strong>: Apache Flink Agents 0.2.0 introduces event-driven AI agents to Flink’s streaming runtime, offering Java API parity with Python, expanded ecosystem integrations, a redesigned memory system, and durable execution for high-volume event streams with millisecond latency. <a href="https://app.daily.dev/posts/ydmla9gYC" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA NVFP4 accelerates AI performance</strong>: NVIDIA introduced NVFP4, a 4-bit floating-point precision format for AI training and inference, achieving up to 15 petaFLOPS on Blackwell Ultra GPUs and maintaining accuracy comparable to higher-precision formats. <a href="https://app.daily.dev/posts/2zfw3Y2wB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Mozilla invests $1 billion in AI infrastructure</strong>: Mozilla announced a $1 billion investment strategy to build AI infrastructure, focusing on foundational rails for trustworthy, smaller-scale AI systems rather than AI models, aiming to challenge model-centric approaches and support regulated industries. <a href="https://app.daily.dev/posts/HZNmJadhL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Svelte’s async-first reactivity challenges react server components</strong>: Svelte’s creator, Rich Harris, highlighted Svelte’s fine-grained reactivity and async-first architecture as an alternative to React Server Components, allowing components to express data dependencies directly and automatically coordinating asynchronous updates. <a href="https://app.daily.dev/feed-by-ids?id=Xqvf8335M&amp;id=s3aE7CClS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI identifies 500 high-severity vulnerabilities, human verification pending</strong>: Anthropic’s Claude Opus 4.6 supposedly identified 500 high-severity vulnerabilities in open-source software autonomously. But don’t get too excited, human verification is ‘underway’ to prevent false positives. Because, you know, AI is great at finding things, not so much at discerning if those things are actually problems or just its own creative interpretations. <a href="https://app.daily.dev/posts/DDZLtuMLV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is taking over open source, or just breaking its monetization model?</strong>: Tailwind CSS laid off 75% of its engineering team, not because it’s unpopular, but because AI tools generate code without developers visiting the documentation site. Apparently, AI is so good it’s disrupting the traditional open-source monetization model that relies on documentation traffic. So, AI isn’t ‘taking over’ open source, it’s just making it harder for humans to get paid for it. <a href="https://app.daily.dev/posts/uR7Q2jMhV" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents built cloud infrastructure in one week, with caveats</strong>: An engineer at Kilo claims to have built Cloud Agents infrastructure in one week using AI agents as ‘team members.’ The secret? Generating onboarding documentation for the AI, maintaining ‘deviation logs,’ and focusing on MVPs. So, AI can build infrastructure fast, as long as you meticulously document everything for it, track its mistakes, and keep it on a very short leash. Sounds like a typical junior engineer, but with more compute. <a href="https://app.daily.dev/posts/vdssMTe1b" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
