<h2>Must know</h2>
<ul>
<li><strong>TSMC sees endless AI chip demand, plans massive 2nm ramp-up</strong>: TSMC reported record Q4 earnings for 2025, with revenue up 25.5% year-over-year, driven by what its CEO calls “endless” AI-related chip demand. The company forecasts 30% revenue growth for 2026 and plans $52-56 billion in capital expenditures, primarily for advanced 7nm and 2nm nodes. TSMC expects its 2nm process to ramp faster than previous generations, with 30% of production slated for the USA, as reported by their official financial statements. <a href="https://app.daily.dev/posts/aTc9ls9yL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical WordPress plugin vulnerability grants unauthenticated admin access</strong>: A critical privilege escalation vulnerability (CVE-2026-23550) in the Modular DS WordPress plugin, affecting versions 2.5.1 and below, allows unauthenticated attackers to gain instant admin access. The flaw, with a CVSS score of 10.0, stems from faulty API routing logic. Active exploitation has been detected, and administrators are urged to update to version 2.5.2 immediately and audit for unauthorized accounts. <a href="https://app.daily.dev/posts/D6wKV3WK1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palo Alto Networks patches high-severity DoS in PAN-OS firewalls</strong>: Palo Alto Networks released patches for CVE-2026-0227, a high-severity denial-of-service vulnerability in PAN-OS firewalls with GlobalProtect enabled. Unauthenticated attackers can exploit this flaw to force firewalls into maintenance mode, potentially causing network outages. Public proof-of-concept code exists, increasing the urgency for manual patching by NGFW users, though Prisma Access customers are largely automatically updated. <a href="https://app.daily.dev/posts/MtreEtTLc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>Databricks and Snowflake face market saturation, pivot to AI</strong>: Databricks and Snowflake are reportedly facing growth saturation in their core data platform markets, prompting both to aggressively pivot into AI and application development to sustain high valuations, with Snowflake actively promoting AI capabilities and developer engagement. <a href="https://app.daily.dev/feed-by-ids?id=Kv2NZR8R1&amp;id=LnOTpvngc" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ClickHouse acquires Langfuse, launches native Postgres service</strong>: ClickHouse, after raising $400M, acquired Langfuse for LLM observability and launched a native Postgres service with Ubicloud, aiming to unify OLTP and OLAP workloads for AI-driven applications. <a href="https://app.daily.dev/feed-by-ids?id=BqVJ6XPkS&amp;id=7DDF3YQda&amp;id=kU2RvmlCw&amp;id=Ycyvr9ct2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Agentic AI memory systems advance with new approaches</strong>: New approaches to AI agent memory, including file-based systems, dynamic tool loading in Claude Code, and Recursive Language Models (RLMs) from MIT, are improving context management and scalability for LLMs. <a href="https://app.daily.dev/feed-by-ids?id=Ks729qZi2&amp;id=SRm19EN1b&amp;id=Y949qVXTN&amp;id=uq0isMFwF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare boosts resilience, acquires Astro Technology Company</strong>: Cloudflare launched ‘Code Orange: Fail Small’ to prevent global outages through phased configuration rollouts and acquired Astro Technology Company, an open-source web framework, which will remain MIT-licensed. <a href="https://app.daily.dev/feed-by-ids?id=OdNHxfcwg&amp;id=Q4Q57JyYL&amp;id=yRAOJZmiR" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Enterprise AI projects stalled by infrastructure complexity and technical debt</strong>: Over half of US enterprises report AI project delays or cancellations due to infrastructure complexity, underutilized GPUs, and technical debt, with MIT research showing 95% of generative AI pilots yield no measurable ROI. <a href="https://app.daily.dev/feed-by-ids?id=k12AAd1Ej&amp;id=tr8ZpGHVk" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>RondoDox botnet exploits HPE OneView, GootLoader evades detection</strong>: The RondoDox botnet is actively exploiting a critical HPE OneView RCE vulnerability (CVE-2025-37164) targeting government organizations, while GootLoader malware uses a novel technique of concatenating thousands of malformed ZIP archives to bypass detection. <a href="https://app.daily.dev/feed-by-ids?id=YZooc8gSw&amp;id=gTlhYMQCn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux kernel enables Intel TSX by default on secure CPUs</strong>: The Linux kernel 7.0 will enable Intel TSX by default on newer CPUs not affected by TAA vulnerabilities, such as Cooper Lake and Ice Lake, aiming to improve performance where safe. <a href="https://app.daily.dev/feed-by-ids?id=kVF0lIVI4&amp;id=l4qet2HfG" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>OpenAI to test targeted ads in ChatGPT: because free AI isn’t free enough</strong>: OpenAI is rolling out targeted ads for free and Go tier ChatGPT users in the U.S., claiming it’s to ‘support continued free access’ and ‘expand affordable AI.’ They assure us ads won’t influence responses and user data won’t be sold, which is exactly what every ad-driven platform says right before it does the opposite. Apparently, losing $74 billion by 2028 isn’t enough motivation to find a business model that doesn’t involve shoving ads into your chatbot conversations. <a href="https://app.daily.dev/feed-by-ids?id=DQh9IAXNl&amp;id=rYkFwNLz3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google’s AI search: ‘unavoidable’ errors, ‘unavoidable’ reputational damage</strong>: Google’s AI search is generating false and defamatory statements about Norwegian politicians, including criminal accusations. Google’s defense? Errors are ‘unavoidable’ given the scale of the internet. So, apparently, it’s unavoidable that their AI will ruin lives, but totally avoidable for Google to take responsibility. Just another day in the ‘move fast and break things, especially reputations’ world of AI. <a href="https://app.daily.dev/feed-by-ids?id=507BvKbPR&amp;id=QiN4er8h3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-powered coding tools: faster, but still needs a human to clean up the mess</strong>: Developers are raving about AI tools like Claude for speeding up coding, cutting a day’s work to an hour. But, surprise, surprise, the AI still can’t handle ‘nuanced software requirements’ like a robust resume function or proper error handling. So, you get your code faster, but you still need a human to babysit the AI and fix its ‘limitations.’ Progress, I guess? <a href="https://app.daily.dev/feed-by-ids?id=SqdiiasJQ&amp;id=YfWBhNQB7&amp;id=H6Mp947gD&amp;id=h6ZJabZ0D&amp;id=laMxMdAYV" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
