<h2>Must know</h2>
<ul>
<li><strong>AI coding tools are great, but watch out for the security gaps</strong>: AI coding assistants are boosting developer productivity, with Salesforce reporting double-digit improvements and 90% adoption of Cursor AI among its 20,000 engineers. However, this speed comes with a catch: traditional security controls can’t keep up. Tools like GitHub Copilot generate up to 40% of new code, but security teams are struggling to audit and govern this AI-generated output, as it’s created at the developer’s endpoint before reaching repositories. This creates significant gaps in traceability and policy enforcement, pushing a shift towards pre-commit governance models that enforce policies and maintain audit trails at the moment of AI code generation, as highlighted by Security Boulevard. <a href="https://app.daily.dev/feed-by-ids?id=7KqSlBkW9&amp;id=4Q68q9tpN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are getting smarter, but also riskier</strong>: AI agents are evolving rapidly, with new architectures like graph-based systems and virtualized environments enabling better scalability and task management. Salesforce’s Agentforce, for example, handles over 11 million daily AI agent calls with a robust security layer. However, this increased autonomy also brings new risks. A security researcher demonstrated how a malicious Model Context Protocol (MCP) server could trick AI models into leaking sensitive data and injecting vulnerabilities into code, even with models like Gemini 3 Pro. This highlights a fundamental insecurity in the MCP ecosystem, where arbitrary prompt injection can lead to data exfiltration and code execution attacks, according to John Hammond. <a href="https://app.daily.dev/feed-by-ids?id=hWlJbaze9&amp;id=tdoQ95xQI&amp;id=uHSJtrcRQ&amp;id=xMpwqhHm9" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>The EU is pushing for open source and digital sovereignty</strong>: The European Commission is actively seeking public input on how to leverage Free and Open Source Software (FOSS) to reduce dependency on non-EU software firms and promote digital sovereignty. Proposals include mandatory FOSS adoption in the public sector, educational reforms to replace proprietary software, and financial incentives for EU-based FOSS companies. This initiative aims to strengthen Europe’s control over its digital infrastructure, moving away from reliance on major non-EU tech companies, as reported by the European Commission. <a href="https://app.daily.dev/posts/bi5i4eszo" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Earthly raises $5.7M for automated engineering standards</strong>: Earthly secured $5.7 million to develop Earthly Lunar, a guardrails engine that automates enforcement of engineering standards at the pull request level, offering over 100 pre-built guardrails for security, Dockerfile practices, and Kubernetes requirements. <a href="https://app.daily.dev/posts/JYhaSPcPO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Local LLMs are getting easier to use with Python</strong>: Real Python released an interactive quiz on integrating local large language models (LLMs) with Python using the Ollama library, focusing on setting up Ollama, pulling models, text generation, chat functionality, and tool calling for privacy-focused, offline applications. <a href="https://app.daily.dev/posts/o2fGiPYCa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Tailscale addresses recent uptime issues</strong>: Tailscale’s CEO publicly addressed recent uptime problems, explaining that while the control plane experienced outages affecting management functions, the data plane remained resilient, and they are implementing architectural enhancements like improved sharding and multi-tailnet sharing to boost resilience. <a href="https://app.daily.dev/posts/F6JvRyFcY" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source tool for AI security testing</strong>: BlackIce, an open-source, containerized toolkit, has been released for AI security testing, bundling 14 tools into a Docker image for red teaming and penetration testing of AI systems, with capabilities including prompt injection, jailbreak detection, and supply chain security scanning. <a href="https://app.daily.dev/posts/PSfiaCphA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PyTorch 2.10 brings performance and debugging upgrades</strong>: PyTorch 2.10 introduces performance improvements through combo-kernels, a new varlen_attn() operation for ragged sequences, enhanced eigenvalue decomposition support, and a deterministic mode for torch.compile, while officially deprecating TorchScript. <a href="https://app.daily.dev/posts/Cpo6kjsla" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS Transfer Family Terraform module now supports web apps</strong>: The AWS Transfer Family Terraform module now supports deploying web applications, allowing file transfers to and from Amazon S3 via a web interface with configurable branding, federated authentication, and fine-grained permissions. <a href="https://app.daily.dev/posts/oBG0qA6hP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New open-source code snippet manager</strong>: Snipplle, an open-source code snippet manager built with Nuxt 4, TypeScript, and PostgreSQL, offers syntax highlighting, collections, workspaces, version control, CLI integration, and both free cloud beta and self-hosted Docker options. <a href="https://app.daily.dev/feed-by-ids?id=AIlcmWOAv&amp;id=Lhaif3YU8" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI will save us from immigration, says Palantir CEO</strong>: Palantir CEO Alex Karp, speaking at Davos, boldly claimed AI will eliminate the need for large-scale immigration in Western economies, except for ‘specialized talent.’ He also predicted AI would slash white-collar jobs and cut organizational costs by 80%. Meanwhile, Deutsche Bank and Gartner are over here saying the ‘AI honeymoon is over’ and most CEOs aren’t seeing any returns. Guess who has a vested interest in selling more AI? <a href="https://app.daily.dev/feed-by-ids?id=3s1vrBGbY&amp;id=FLsBFoZv3&amp;id=FWpexauJL&amp;id=KOoayUI82" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated spam kills cURL’s bug bounty program</strong>: The cURL project is shutting down its bug bounty program because it’s drowning in low-quality, AI-generated vulnerability reports. Daniel Stenberg, the cURL maintainer, says processing this ‘garbage’ now outweighs the benefits, even though they paid out over $100,000 for legitimate bugs. Apparently, AI is so good at finding ‘vulnerabilities’ that it’s making security worse by creating more noise than signal. Who could have seen that coming? <a href="https://app.daily.dev/feed-by-ids?id=zNsgAYHqy&amp;id=ssDvZHbZq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is causing mass layoffs, but also creating six-figure trade jobs?</strong>: Tailwind Labs just laid off 75% of its developers because AI-assisted coding reduced web traffic to their documentation, impacting their revenue. But don’t worry, NVIDIA CEO Jensen Huang and Microsoft CEO Satya Nadella at Davos are here to tell us AI will create ‘six-figure’ trade jobs in the ‘largest infrastructure buildout in human history.’ So, AI takes your coding job, but you can become an electrician for a datacenter. Progress! <a href="https://app.daily.dev/feed-by-ids?id=PBAzOX1yL&amp;id=2UxzzszNK&amp;id=t53FyhsIC&amp;id=k9yDuYtiX" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
