<h2>Must know</h2>
<ul>
<li><strong>John Schulman on the engineering shift in AI research</strong>: OpenAI co-founder John Schulman, now at Thinking Machines, highlights a significant industry pivot where talent distribution is moving from ‘research taste’ toward heavy engineering. In a deep dive with Cursor, he discusses the future of co-training generators and verifiers, while introducing Tinker, a low-level fine-tuning API designed to let researchers run post-training algorithms without the overhead of GPU infrastructure management. He also reflects on the ‘ragtag’ origins of OpenAI and the missed opportunities that could have brought ChatGPT to market even earlier. <a href="https://app.daily.dev/posts/7Zbxx5IYi" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The accelerating vendorization of the software stack</strong>: The industry is undergoing a fundamental shift from in-house development to purchasing specialized vendor solutions for everything from observability to AI APIs. According to analysis by Erik Bernhardsson, this ‘vendorization’ explains recent big-tech layoffs and the rising salaries at vendor companies where engineers have outsized ecosystem impact. This trend is further accelerated by high GPU costs, which favor multi-tenant API models over self-hosted open-source alternatives. <a href="https://app.daily.dev/posts/a5rbOtmsv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS solves the 18 year old S3 bucket naming headache</strong>: AWS has finally introduced account-regional namespaces for S3, effectively ending the global bucket name collision issue that has plagued developers since 2006. This opt-in change allows bucket names to be scoped to specific accounts and regions, eliminating <code>BucketAlreadyExists</code> errors in IaC pipelines and mitigating ‘confused-deputy’ security risks. The update is backward-compatible and available across 35 regions at no additional cost. <a href="https://app.daily.dev/posts/8bZsUijKD" 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</strong>: The major release adds first-class optimistic updates, a new <code>useHttp</code> hook, and SSR support in Vite dev mode without a separate Node process. <a href="https://app.daily.dev/posts/WWmxp0DEw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Materialize achieves 100x speedup in Timely Dataflow</strong>: By implementing a ‘notify only if holding a capability’ scheduling mode, operator visits were reduced from one million to just one thousand per iteration. <a href="https://app.daily.dev/posts/6leNCC2Gz" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Amazon Aurora PostgreSQL now supports two click serverless creation</strong>: A new express configuration enables sub-second database setup with IAM-based passwordless authentication and Vercel integration. <a href="https://app.daily.dev/posts/MjUWMvC5e" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Ubuntu switches to Rust based NTP implementation</strong>: Canonical plans to replace Chrony and LinuxPTP with <code>ntpd-rs</code> starting in Ubuntu 26.10 to improve memory safety and security. <a href="https://app.daily.dev/posts/MuH00Ge5n" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Snowflake launches Batch Cortex Search for high throughput hybrid queries</strong>: The new function is reportedly 47.5x faster than standard Cortex Search for large-scale entity resolution and content tagging tasks. <a href="https://app.daily.dev/posts/D2fJfZuGA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Security as Code becomes the baseline for continuous compliance</strong>: Driven by the EU Cyber Resilience Act, organizations are moving away from periodic audits toward automated policy enforcement via OPA and Checkov. <a href="https://app.daily.dev/posts/kjzObhPK3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Appoval automates iOS App Store guideline compliance checks</strong>: The tool uses AI to scan projects for common rejection triggers like missing privacy manifests or entitlement mismatches before submission. <a href="https://app.daily.dev/posts/sm9ro8NNa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>The AI job search arms race is officially here</strong>: One developer built a system to scrape LinkedIn, run ‘strengths-and-gaps’ analysis, and auto-generate ‘tailored’ resumes and cover letters using a fleet of LLMs. We’ve reached the point where AI-generated applications are being screened by AI-powered ATS systems, creating a perfectly efficient loop of synthetic nonsense that helps absolutely no one. It’s not ‘self-aware candidacy’; it’s just DDoS-ing the hiring process with high-quality spam. <a href="https://app.daily.dev/posts/0YroeqW2V" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI didn’t break your culture, it just showed how lazy it was</strong>: Engineering teams are currently panicking that AI tools are producing ‘junk drawer’ codebases, but the reality is they were already deferring to ‘oracles’ like Stack Overflow or senior devs without understanding the ‘why.’ If your team’s architectural integrity collapses because a chatbot gave you a hallucinated library, you didn’t have a culture of judgment; you had a culture of copy-pasting. AI is just the mirror showing us the technical debt we were already happy to ignore. <a href="https://app.daily.dev/feed-by-ids?id=eYrc9sF1z&amp;id=p8Xi3BQzS" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The environmental rebound effect of ‘efficient’ AI</strong>: While Google and others tout ‘TurboQuant’ and model compression as wins for the planet, history shows that efficiency gains almost always lead to expanded usage rather than reduced footprints. We are celebrating 6x memory reductions while simultaneously scaling inference volumes by 100x, effectively treating sustainability as a PR slide rather than a hard design constraint. It’s the Jevons Paradox in a hoodie: the more efficient we make the GPUs, the more compute we’ll find a way to waste. <a href="https://app.daily.dev/feed-by-ids?id=yBDFJf8H2&amp;id=PTvjeLzdh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
