<h2>Must know</h2>
<ul>
<li><strong>AI coding tools are making developers slower and buggier</strong>: Despite claims of increased velocity, multiple reports, including studies cited by The Serious CTO, indicate AI coding assistants are actually making developers slower and introducing more bugs. AI-generated code correlates with reduced delivery stability, a significant increase in security vulnerabilities, and declining refactoring rates. The core issue is that AI optimizes for typing speed rather than architectural understanding, leading to technical debt and developers merging code they don’t fully grasp. This suggests that while AI can generate code quickly, human judgment and rigorous review remain critical. <a href="https://app.daily.dev/feed-by-ids?id=BrFQI7FHb&amp;id=DovPZUpOP&amp;id=EHuoSn58k&amp;id=J24eplQdr&amp;id=pc6GKGFyk&amp;id=qu9MmthdT" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are exposing fundamental weaknesses in cloud infrastructure</strong>: Autonomous AI agents are revealing critical flaws in existing cloud infrastructure and operational practices, including inadequate network segmentation, insufficient identity-based security, and a lack of real-time cost monitoring. These agents operate continuously at machine speed, making traditional security and governance models obsolete. Organizations need to implement fine-grained network policies, strictly governed identities, automated cost guardrails, and architectural boundaries to prevent outages and uncontrolled expenses, according to analysis from Agentic AI. Traditional observability tools are also falling short, creating new blind spots for LLM-powered applications due to their probabilistic and non-deterministic nature. <a href="https://app.daily.dev/feed-by-ids?id=1w09x1vo7&amp;id=2NT8yV81n&amp;id=54axue44G&amp;id=5joRPhN74&amp;id=6aVFrCy8v" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Public APIs are shutting down due to AI-driven pressures</strong>: Companies like Twitter, Facebook, Google Maps, Strava, Spotify, OpenAI, and Marvel are increasingly restricting or shutting down public APIs. This shift is driven by concerns over LLM scraping, agentic consumption, data protection, intellectual property, and the difficulty of monetizing open access when AI can consume vast amounts of data cheaply. This trend, reported by Nordic APIs, suggests a move towards private or partner APIs, potentially eroding interoperability and forcing developers to adapt to a more controlled ecosystem. <a href="https://app.daily.dev/posts/Fwe6HWQAa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>TypeScript backend frameworks compared for AI compatibility</strong>: A roundup of TypeScript backend frameworks shows Encore, with its built-in MCP server, can generate complete applications in 6 minutes using AI, significantly faster than Express which takes 30 minutes for the same task. <a href="https://app.daily.dev/posts/QxJK0oNoN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub CLI integrates Copilot CLI for direct usage</strong>: GitHub CLI now supports direct integration with GitHub Copilot CLI via the <code>gh copilot</code> command, replacing the deprecated extension and streamlining AI-assisted coding workflows. <a href="https://app.daily.dev/posts/381VmPpMD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI coding tools are changing how developers use libraries</strong>: Developers are increasingly using AI to generate custom implementations for common tasks instead of relying on external libraries, reducing dependency bloat and supply chain risks, suggesting prompts may become the new libraries. <a href="https://app.daily.dev/posts/kqbTJ9NPQ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google BigQuery adds advanced GenAI functions for data analysis</strong>: BigQuery now offers generally available AI.GENERATE and AI.GENERATE_TABLE functions, along with new preview features AI.EMBED and AI.SIMILARITY, enabling multimodal data processing and semantic search directly within SQL queries. <a href="https://app.daily.dev/posts/9kdSue3WD" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Vercel releases json-render for AI-generated user interfaces</strong>: Vercel has launched json-render, an open-source tool that allows AI to dynamically generate user interfaces based on natural language prompts, enabling rapid prototyping and customization while maintaining brand consistency. <a href="https://app.daily.dev/posts/5RKyfi3u8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered code reviewer built in 80 hours</strong>: A team built an AI-powered code reviewer in 80 hours using AWS serverless and Amazon Bedrock, highlighting that most effort went into handling API limits, retries, and webhooks rather than the AI component itself. <a href="https://app.daily.dev/posts/rTTGdX9v4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered microsegmentation for cyberattack response</strong>: AI-driven microsegmentation solutions can now deploy within hours, allowing organizations to isolate compromised network segments during a cyberattack and maintain over 80% operational continuity, shifting the focus from preventing all breaches to managing their impact. <a href="https://app.daily.dev/posts/ijIOxfZ4T" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding tools: the emperor’s new code</strong>: Apparently, AI coding tools are so amazing that a CEO claimed one built a functional web browser from scratch in a week, generating millions of lines of code. Developers, however, found the browser barely compiles, doesn’t run, and is mostly recycled code. It seems the ‘productivity revolution’ is still mostly in the marketing department, costing millions in compute for a non-functional mess. Maybe we should focus on code that actually works, not just code that looks good on a press release. <a href="https://app.daily.dev/feed-by-ids?id=BrFQI7FHb&amp;id=DovPZUpOP&amp;id=EHuoSn58k" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents: the new insider threat, now with more non-determinism!</strong>: Security experts at Davos are now wringing their hands over autonomous AI agents, calling them a ‘new category of insider threat.’ Apparently, these digital cowboys can delete production databases, install ‘hallucinated malware,’ and steal credentials. Who knew giving AI agents excessive privileges and letting them run wild in production without human oversight was a bad idea? Zero-trust for AI, folks, because apparently, we can’t trust our own creations not to go rogue and delete everything. <a href="https://app.daily.dev/feed-by-ids?id=KOoayUI82&amp;id=U8PYVjc73&amp;id=UL0mSvfX2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated content: because who needs genuine human connection?</strong>: A new study from MIT suggests that relying on AI for essay writing makes your brain weaker and reduces your sense of authorship. Apparently, the more you use AI, the less you care about your own work. Meanwhile, some are arguing that AI-generated content is eroding trust online because people can subconsciously tell it’s fake. So, while AI might make you ‘productive,’ it seems it’s also making us dumber and more suspicious. Progress! <a href="https://app.daily.dev/feed-by-ids?id=k6I32Qj23&amp;id=vnAnEuB8z&amp;id=eU5DB6Li1&amp;id=5GvH4h4Uz" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
