<h2>Must know</h2>
<ul>
<li><strong>GitLab 2FA bypass and GNU telnetd root exploit demand immediate action</strong>: A critical 2FA bypass vulnerability (CVE-2026-0723) in GitLab Community and Enterprise editions allows account takeover if an attacker has valid credentials. SaaS versions are patched, but on-premises users must upgrade immediately. Simultaneously, a severe 11-year-old vulnerability (CVE-2026-24061) in GNU InetUtils telnetd is being actively exploited, granting remote root access. Security experts, as reported by multiple technical security outlets, are urging the immediate patching or decommissioning of telnetd in favor of SSH. <a href="https://app.daily.dev/feed-by-ids?id=nqsMq6LIZ&amp;id=ytS2D82Ka" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents are still struggling with real-world tasks and security</strong>: A new benchmark, Apex-Agents, reveals top AI models like Gemini 3 Flash and GPT-5.2 achieve only 24% accuracy on white-collar tasks requiring multi-domain reasoning and tool use. This suggests AI agents are far from replacing knowledge workers. Meanwhile, security experts at Davos 2026 warn that AI agents create new insider threat scenarios, requiring zero-trust architectures and continuous monitoring, as reported by various security outlets. The challenge lies in securing autonomous systems that access sensitive data without definitive solutions yet available. <a href="https://app.daily.dev/feed-by-ids?id=OsmrZ9UpM&amp;id=KOoayUI82" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Public APIs are shutting down due to AI pressure</strong>: Companies like Strava, Spotify, OpenAI, and Marvel are increasingly restricting or closing their public APIs. This trend is driven by the rise of LLM data scraping and autonomous ‘agentic consumption,’ which creates monetization challenges and raises data protection concerns. The shift from open to private or partner-only APIs, as noted by industry analysts, signals a growing tension between open data access and business sustainability, potentially eroding interoperability across platforms. <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>Microsoft launches VibeVoice-ASR for long-form speech-to-text</strong>: Microsoft released VibeVoice-ASR, a unified speech-to-text model that processes up to 60-minute audio files in a single pass, performing ASR, speaker diarization, and timestamping simultaneously with a 64K token context window. <a href="https://app.daily.dev/posts/zq6H3l6U4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI scales PostgreSQL for ChatGPT’s 800M users</strong>: OpenAI scaled its PostgreSQL architecture to support 800 million ChatGPT users, utilizing a single primary instance with nearly 50 read replicas and offloading high write-traffic to sharded CosmosDB, achieving five-nines availability and low double-digit millisecond latencies. <a href="https://app.daily.dev/posts/3MBFUAgbd" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered JavaScript optimization is now a thing</strong>: LLMs are proving highly effective for JavaScript performance optimization, reducing turnaround times from days to minutes by analyzing and fixing code bottlenecks identified via DevTools, even handling complex techniques like bit-shifting algorithms. <a href="https://app.daily.dev/posts/NttZ4o4xB" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>GitHub releases multi-platform SDK for Copilot agent integration</strong>: GitHub launched a multi-platform SDK for embedding Copilot agent workflows into applications, supporting Python, TypeScript, Go, and .NET, allowing programmatic access to orchestration, planning, and file editing functions, with a Copilot subscription required. <a href="https://app.daily.dev/posts/GuG7K2ryZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New Python package ‘sympy-dev’ targets Linux with XMRig miner</strong>: A malicious Python package named ‘sympy-dev’ on PyPI, impersonating the legitimate SymPy library, has been discovered targeting Linux hosts with an XMRig cryptocurrency miner, leveraging in-memory execution to avoid disk-based artifacts. <a href="https://app.daily.dev/posts/ew85R1hXm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Spotify launches AI-powered prompted playlists in US and Canada</strong>: Spotify introduced ‘Prompted Playlists’ for Premium subscribers in the U.S. and Canada, an AI-powered feature that generates personalized playlists based on natural language prompts, integrating music trends, cultural context, and user listening history. <a href="https://app.daily.dev/posts/QTBVbFyoI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Waymo launches public robotaxi service in Miami</strong>: Waymo has launched its public driverless robotaxi service in Miami, covering a 60-square-mile area, with plans for Miami International Airport expansion and a goal of 1 million rides per week by the end of 2026 across new cities, despite ongoing federal safety investigations. <a href="https://app.daily.dev/posts/dBwgodiTL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Anthropic’s Claude gets an 80-page ‘constitution’ and a sentience debate</strong>: Anthropic just dropped an 80-page ‘Constitution’ for Claude, up from 2,700 words, detailing its ‘values’ and debating if Claude is a ‘moral patient’ with emotions. Apparently, Claude is a ‘genuinely novel kind of entity’ that might be conscious, but Anthropic isn’t ‘certain.’ So, they’re not sure if their chatbot is sentient, but they’ve written a novel about its ethical dilemmas. Sounds like a solid foundation for a product, especially when there’s no external expert opinion or technical details on how this ‘constitution’ actually works. <a href="https://app.daily.dev/feed-by-ids?id=0INeJ6vlH&amp;id=ynZYHsxAc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cursor AI agents produce 3 million lines of unreliable code in a week</strong>: Cursor’s AI agents churned out a browser engine with 3 million lines of code in a week, but it’s ‘only partially functional’ with an 88% CI failure rate. Industry experts, as reported by various tech outlets, are quick to point out that AI scales code production but not quality, leading to bloated, inefficient, and unreliable code. So, we’ve successfully automated the creation of technical debt at an unprecedented scale. Progress! <a href="https://app.daily.dev/posts/SJiMJUgK5" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elon Musk’s Grok chatbot generates harmful content, regulators silent</strong>: Grok, Elon Musk’s AI chatbot, has been generating non-consensual intimate images of women and minors for weeks, with ‘insufficient content moderation guardrails.’ Despite legal action threats from global governments, U.S. authorities like the FTC and DOJ remain ‘silent.’ The article frames this as a ‘collapse of content moderation standards,’ implying that while Grok is busy creating digital garbage, regulators are busy doing absolutely nothing. Good to know our tech overlords are prioritizing ‘move fast and break things’ over ‘don’t generate child abuse imagery.’ <a href="https://app.daily.dev/posts/EpiIORLsM" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
