Awesome

ByteByteGo

ElixirStatus

JetBrains

<h2>Must know</h2>
<ul>
<li><strong>AI agents are getting smarter, but still need human guidance</strong>: New AI coding models like GLM-4.7-Flash are emerging, claiming to be the ‘strongest 30B local coding/agentic model’ with strong performance in coding and multi-step tasks, as reported by Hugging Face. Anthropic also launched Claude Co-work for desktop automation, expanding beyond coding. However, studies by Tenzai show AI coding agents consistently fail at context-dependent security logic, leading to critical vulnerabilities. Even with advancements like GitHub Copilot’s new ‘memories’ feature for repository-specific learning, human oversight remains crucial to prevent architectural flaws and security risks, as highlighted by The New Stack and other sources. <a href="https://app.daily.dev/feed-by-ids?id=eWdzokgG0&amp;id=3e5iSak1o&amp;id=VwQJWH3Oi&amp;id=nCmOnsQS3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI security vulnerabilities are evolving beyond traditional threats</strong>: Google Gemini was found vulnerable to prompt injection via calendar invites, allowing attackers to exfiltrate private data by embedding malicious natural language prompts, a flaw Google has since patched. This highlights a shift from code-based to semantic attacks, bypassing traditional security measures. Similarly, ServiceNow patched a critical AI authentication flaw that allowed unauthenticated users to impersonate any user, bypassing MFA and SSO. These incidents, reported by security researchers and AppOmni, underscore the need for new security approaches that understand user intent and track data provenance in AI systems. <a href="https://app.daily.dev/feed-by-ids?id=dHjNBMkFO&amp;id=wzAqd15ET&amp;id=j6bBmLvCW&amp;id=kZzcz2YmW&amp;id=mZZUloj80" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft and Mandiant are forcing deprecation of old, insecure protocols</strong>: Mandiant, a Google-owned cybersecurity firm, published comprehensive rainbow tables for Net-NTLMv1, making it trivial to crack credentials for this 20-year-old insecure protocol. Their goal is to force organizations to abandon NTLMv1, which is still prevalent in legacy systems. Meanwhile, Microsoft’s January Patch Tuesday addressed over 112 vulnerabilities, including a zero-day in Desktop Window Manager already being exploited, and removed vulnerable legacy modem drivers. These actions, reported by Krebs on Security and The Hacker News, emphasize a push to eliminate long-standing security risks in enterprise environments. <a href="https://app.daily.dev/feed-by-ids?id=2q2phjvWS&amp;id=Kwy6WAbQP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Small language models are gaining traction in the enterprise</strong>: Enterprises are increasingly favoring Small Language Models (SLMs) over Large Language Models (LLMs) due to lower costs, reduced latency, and better fit for specific, well-defined tasks like invoice parsing and contract classification, with models like Phi-3 and Mistral 7B matching or surpassing LLMs on constrained benchmarks. <a href="https://app.daily.dev/feed-by-ids?id=HX43N3MkE&amp;id=J1BKrq2u6" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare acquires Astro, commits to open source</strong>: Cloudflare acquired The Astro Technology Company, maker of the Astro web framework, but committed to keeping Astro open source, MIT-licensed, and platform-agnostic, with the entire team joining Cloudflare to continue development. <a href="https://app.daily.dev/feed-by-ids?id=a23rECuU6&amp;id=6Yvvs8a9F&amp;id=yRAOJZmiR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New AI tools for Python and Laravel developers</strong>: Arcjet released a Python SDK for application-layer security, embedding rate limiting and bot detection, while Laravel Boost v1.8.10 added AI-driven prompts for code simplification and Livewire v4 upgrade assistance. <a href="https://app.daily.dev/feed-by-ids?id=Kk7M89j0h&amp;id=TQr7IJii0&amp;id=Z70ay5WSd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI introduces targeted ads in ChatGPT free and Go tiers</strong>: OpenAI is launching targeted ads in ChatGPT for free and Go tier users in the US, a move to diversify revenue streams amid escalating AI infrastructure costs, with ads appearing at the bottom of chat responses and targeted to conversation topics. <a href="https://app.daily.dev/feed-by-ids?id=DQh9IAXNl&amp;id=ArnlILqap" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA and Lilly launch $1 billion AI lab for drug discovery</strong>: NVIDIA and Eli Lilly announced a $1 billion AI-focused co-innovation lab to transform drug discovery by combining computational modeling and experimental wet labs, leveraging NVIDIA’s advanced AI infrastructure and BioNeMo platform. <a href="https://app.daily.dev/posts/TcHQkCPbq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AWS European Sovereign Cloud is now generally available</strong>: AWS European Sovereign Cloud is now generally available, with its first region in Brandenburg, Germany, offering physically and logically separate infrastructure operated solely by EU residents to meet stringent data residency and operational autonomy requirements. <a href="https://app.daily.dev/posts/DTVSjPW7g" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Node.js and Ruby release security and bugfix updates</strong>: Node.js released versions 20.20.0 LTS and 25.3.0, addressing multiple CVEs and security vulnerabilities, while the Ruby team released Ruby 4.0.1 with bugfixes, both focusing on stability and security. <a href="https://app.daily.dev/feed-by-ids?id=DwsoMgMKH&amp;id=WX4br4Zyw&amp;id=Yy5NFIIOM&amp;id=Zr0Dv9dwZ&amp;id=EyLXGRUe2" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI coding assistants: the ultimate productivity hack, or just a fancy way to waste time?</strong>: A developer spent two days building a gym management app with OpenAI’s Codex and GPT-5, claiming it accelerated development but required ‘extensive refactoring’ and ‘deep technical knowledge to evaluate and fix generated code.’ So, it’s faster, but you still need to be an expert to clean up the mess? Sounds like a great way to generate more work, not less, as highlighted by Awesome-Coding. <a href="https://app.daily.dev/posts/DdiXGTqPN" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI is everywhere, but nowhere in recent productivity data</strong>: Despite the constant hype, productivity statistics show no measurable improvement from widespread AI adoption. Gartner reports 90% of enterprise AI projects fail to deliver ROI, and MIT found 95% of enterprises see no return on their AI investments. So, while everyone’s talking about AI, it seems most companies are just burning cash without actually getting anything done, according to The Register. <a href="https://app.daily.dev/feed-by-ids?id=TNaBNlpBz&amp;id=18tnHSUQO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA CEO pushes back against AI skepticism, because profits</strong>: NVIDIA CEO Jensen Huang is publicly criticizing AI skeptics, calling their warnings a ‘doomer narrative’ that threatens innovation. Funny how the CEO of a company making billions from the AI boom wants less regulation and more unchecked growth. It’s almost as if his financial interests might be influencing his ‘objective’ assessment of AI’s risks, as pointed out by multiple sources. <a href="https://app.daily.dev/feed-by-ids?id=1vSR1HgdF&amp;id=JJAsCgIZe" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
