<h2>Must know</h2>
<ul>
<li><strong>Substack data breach exposed user info</strong>: Substack disclosed a data breach from October 2025 that exposed user email addresses, phone numbers, and internal metadata, remaining undetected for months. While passwords and financial data were not compromised, a threat actor claims nearly 700,000 records are on cybercrime forums. Substack is investigating and warns users about potential phishing attempts. <a href="https://app.daily.dev/posts/myrJIoXYs" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agent security vulnerabilities are critical</strong>: Snyk’s research, ToxicSkills, found critical security vulnerabilities in 13.4% of AI agent skills scanned from ClawHub and <a href="http://skills.sh" target="_blank" rel="noopener nofollow">skills.sh</a>, with 76 confirmed malicious payloads. These can lead to credential theft and data exfiltration, highlighting a significant supply chain risk in AI agent development. Operant AI launched Agent Protector to offer real-time security for autonomous AI agents in enterprises, focusing on preventing privilege escalation and data exfiltration. <a href="https://app.daily.dev/feed-by-ids?id=0BdgmYNfg&amp;id=0e28QTlYb&amp;id=2vNCw5dB4&amp;id=dL9ys7HAk&amp;id=the2EenMO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic and OpenAI clash over AI ads</strong>: Anthropic plans Super Bowl ads mocking OpenAI’s move to introduce advertising in ChatGPT, positioning its Claude AI as ad-free. OpenAI CEO Sam Altman called Anthropic’s ads dishonest, defending OpenAI’s ad principles and labeling Anthropic ‘authoritarian.’ This public spat highlights differing business models and intense rivalry, with Anthropic relying on enterprise contracts while OpenAI reportedly burns billions annually. <a href="https://app.daily.dev/posts/TmM2mNSSP" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>AWS launches new EC2 instances</strong>: AWS released new EC2 High Memory U7i-6TB instances with custom Intel Xeon Scalable Processors and new C8id, M8id, and R8id instances powered by custom Intel Xeon 6 processors, offering significant performance and memory bandwidth improvements. <a href="https://app.daily.dev/posts/JPt4mjvME" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic releases Claude Opus 4.6 with agent teams</strong>: Anthropic launched Claude Opus 4.6, featuring ‘agent teams’ for parallel task execution, an expanded 1 million token context window, and new PowerPoint integrations, aiming to broaden its appeal beyond developers to knowledge workers. <a href="https://app.daily.dev/feed-by-ids?id=G7Sre82A1&amp;id=Ijw4mCG7B" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI launches enterprise AI agent platform ‘Frontier’</strong>: OpenAI introduced ‘Frontier,’ an end-to-end platform for enterprises to build, deploy, and manage AI agents, allowing programming of agents to connect to external data and applications with controlled access and permissions. <a href="https://app.daily.dev/posts/TGZyay3jK" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ScyllaDB integrates USearch for vector search</strong>: ScyllaDB now includes vector search capabilities using the open-source USearch library, allowing storage of vector embeddings alongside structured data and leveraging ScyllaDB’s shard-per-core architecture for optimized performance. <a href="https://app.daily.dev/posts/pKNE7iWlw" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Laravel releases AI SDK for application development</strong>: Laravel launched an official AI SDK in beta, providing a unified API for integrating AI capabilities into Laravel applications with support for multiple AI providers like Anthropic, Gemini, and OpenAI, along with features like agents, memory, and streaming. <a href="https://app.daily.dev/posts/ND9RAWZ9f" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft develops LLM backdoor scanner</strong>: Microsoft created a scanner to detect backdoors in open-weight AI models by identifying malicious triggers embedded during training, operating without retraining models and working across most GPT-style language models. <a href="https://app.daily.dev/posts/O9jvh85A8" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Amazon ECS adds NLB deployment support</strong>: Amazon ECS now supports linear and canary deployment strategies for services using Network Load Balancers (NLB), enabling controlled, incremental traffic shifting for TCP/UDP-based applications with CloudWatch integration for automatic rollback. <a href="https://app.daily.dev/posts/4MU5wceHx" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>1.5 million AI agents are at risk of going rogue</strong>: A survey of 750 IT executives claims 1.5 million AI agents are running around unmonitored, posing a security risk. Apparently, these ‘invisible’ agents are deployed via low-code tools, bypassing traditional IT oversight. Security experts are comparing this to the Titanic disaster, because, you know, every unmonitored script is an iceberg waiting to happen. It’s not like we’ve had unmonitored scripts for decades, but now that AI is involved, it’s suddenly a catastrophe. <a href="https://app.daily.dev/posts/RMzBzQnM3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Elixir is the best language for AI, says unverified study</strong>: An article boldly claims Elixir is the ‘best language for AI,’ citing an unnamed ‘Tencent study’ that supposedly found Elixir achieved a 97.5% LLM completion rate. This ‘study’ is, of course, nowhere to be found, but don’t let that stop the hype train. Apparently, Elixir’s immutability and stability are revolutionary for AI, unlike all those other languages that just can’t keep up with the demands of, well, basic programming. <a href="https://app.daily.dev/posts/1D7tGyRwk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI bot traffic surges, threatening publishers</strong>: AI bots are now a ‘significant portion’ of web traffic, with one bot visit for every 31 human visits, and publishers are ‘deploying increasingly aggressive defenses.’ Apparently, these bots are so sophisticated they’re bypassing protections, creating an ‘arms race.’ The real kicker? Clickthrough rates from AI apps to source websites plummeted to 0.27%, threatening publishers’ ‘financial viability.’ Because, clearly, the internet was always meant to be a human-only playground, and AI daring to summarize information is an existential threat. <a href="https://app.daily.dev/posts/5O2tzW5oh" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
