<h2>Must know</h2>
<ul>
<li>AI agents are transforming software development: AI agents are fundamentally changing how software is built, moving beyond simple code generation to autonomous development. StrongDM, for instance, implemented a “Software Factory” where AI agents write, test, and validate code without human intervention, even cloning third-party services for validation. This shift means developers will focus more on architecture and design, with AI handling the grunt work, as highlighted by StrongDM’s goal to spend $1,000 per day per engineer on AI tokens. <a href="https://app.daily.dev/feed-by-ids?id=HjzMVzPQS&amp;id=SVJSQF1SS&amp;id=QeCjGcTtT&amp;id=vQjfzdkpo" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>The hidden costs of microservices: A 15-engineer company saw a 560% infrastructure cost increase and a 60% productivity drop after decomposing its monolith into 12 microservices, leading to three senior engineers quitting. This illustrates the significant hidden costs of distributed architectures, which often multiply complexity and operational overhead. Companies like Shopify and Stack Overflow demonstrate that well-designed monoliths can scale effectively, suggesting microservices only yield positive returns for organizations with substantial revenue and engineering teams, according to CodeGood. <a href="https://app.daily.dev/posts/L8YCqEU1k" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Cloudflare’s DNS outage and AI agent platform: Cloudflare experienced a DNS outage due to a memory optimization that incorrectly reordered CNAME records, affecting clients expecting a specific order. This highlights the fragility of core internet infrastructure. Separately, Cloudflare introduced Moltworker, an open-source platform for running self-hosted AI agents on its Developer Platform, leveraging Workers, R2, and AI Gateway for secure, scalable deployment at the edge, though it raises questions about local control. <a href="https://app.daily.dev/feed-by-ids?id=SKYTauJXO&amp;id=TpsLSMM7l" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li>AI’s impact on developer careers and code quality: AI is reducing the cost of writing business logic to near-zero, forcing developers to adapt by focusing on system architecture and problem-solving, or risk diminishing economic value, while also raising concerns about code quality and the loss of coding as a craft. <a href="https://app.daily.dev/feed-by-ids?id=1kY4bQ43R&amp;id=EfUZYBNep&amp;id=RhDtuknZS&amp;id=aREPiFEe8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI agent integration and governance: Control planes and automation layers are becoming incident sources, with AI agents like GitHub’s Agent HQ requiring identity, permissions, and guardrails, while prompt injection risks like DockerDash highlight the need for robust governance for Model Context Protocols (MCPs). <a href="https://app.daily.dev/feed-by-ids?id=89Ue9XF5U&amp;id=DDF1LGNa7&amp;id=FGuAm9dTJ&amp;id=NnOA92bnL&amp;id=WndaKmd4O&amp;id=h2Td8u8vN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI privacy and reliability concerns: Incognito mode doesn’t protect data with AI chatbots, as conversations are used for training, and telecom partnerships offering free AI access are designed for data collection, necessitating mathematical redaction or verification systems to prevent hallucinations in critical applications. <a href="https://app.daily.dev/feed-by-ids?id=omgT5xY4B&amp;id=pVBEuoaIJ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Docker vs Nix for reproducibility: Docker containers aren’t truly reproducible due to non-deterministic builds, while Nix offers genuine reproducibility through sandboxed builds, cryptographic hashing, and deterministic environments, albeit with a steeper learning curve. <a href="https://app.daily.dev/posts/R1CRTUmLA" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>New Rails workflow engine geneva_drive: Geneva_drive is a new Rails workflow engine that treats workflows as ActiveRecord models, offering durable execution with queryable state, built-in retries, and arbitrary wait times without external dependencies, simplifying complex workflow management. <a href="https://app.daily.dev/posts/LHscy0nsU" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>UniFi firewalls get major 2025 update: UniFi Network application versions 9 through 9.5 introduced significant firewall improvements, including zone-based firewalls, enhanced policy management, high availability, and AI-powered port anomaly detection, making the platform more robust for production environments. <a href="https://app.daily.dev/posts/YIx2m6Lku" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>CodeQL 2.24.1 improves Maven and query accuracy: CodeQL 2.24.1 enhances Maven private registry support, adds Kotlin 2.3.0, and improves query accuracy for C/C++ buffer overflows and Java lock analysis, while also introducing an experimental prompt injection detection query for LLM code. <a href="https://app.daily.dev/posts/t6yIEU56N" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li>AI-assisted coding is killing open source: Apparently, “vibe coding” (AI-assisted coding) is eroding codebase familiarity and turning maintainers into prompt engineers, threatening open source sustainability by bypassing traditional contribution models. Because, you know, the real problem with open source was always too much human creativity and not enough AI-generated boilerplate. <a href="https://app.daily.dev/feed-by-ids?id=0QENbqZDa&amp;id=lGQ4yZr5a" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>AI solves LEGO’s hardest engineering problem: Carnegie Mellon researchers developed Brick GPT, an AI system that can generate buildable, structurally stable LEGO models from text prompts, solving a problem that supposedly took 40 years. Because, clearly, the biggest challenge in engineering was figuring out how to stack plastic bricks, and only AI could save us from the combinatorial explosion of 915 million combinations for six 2x4 bricks. <a href="https://app.daily.dev/posts/m2dPaArWb" target="_blank" rel="noopener nofollow">Read more</a></li>
<li>Haskell for all: beyond agentic coding: Some folks are claiming that agentic coding tools with chat interfaces break developer flow state and reduce productivity, despite generating code quickly. They suggest “calm technology” like inline suggestions is better. Because, obviously, the problem isn’t the tools themselves, but the fact that developers might actually have to think and interact with something other than their IDE for a second. <a href="https://app.daily.dev/posts/wvSFqs26P" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>