<h2>Must know</h2>
<ul>
<li><strong>AI agent architectures get smarter with memory and context</strong>: The AI agent ecosystem is rapidly evolving with new memory and context management systems. Harrison Chase’s Agent Builder uses file-based memory in Postgres, accessible as a virtual filesystem, for no-code AI agents. Claude Code now dynamically loads relevant tools, reducing context window usage by up to 95%, as noted by Claude Code’s updates. MIT researchers are also pushing Recursive Language Models (RLMs) that can process over 10 million tokens by storing context in modular pieces, making AI agents more efficient and scalable, according to MIT Technology Review. <a href="https://app.daily.dev/feed-by-ids?id=Ks729qZi2&amp;id=SRm19EN1b&amp;id=Y949qVXTN&amp;id=uq0isMFwF&amp;id=WwZicivI3&amp;id=WyRfCzQm3&amp;id=Y9B2mh3yd&amp;id=hx43N3MkE&amp;id=ETkKEZbZ4" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI and Microsoft face jury trial over nonprofit mission</strong>: Elon Musk’s lawsuit against OpenAI and Microsoft, alleging a breach of OpenAI’s founding nonprofit mission, is heading to a jury trial. A federal judge rejected motions to dismiss from both companies, meaning a jury will now decide if OpenAI prioritized profits over its original commitments and if Microsoft aided in that alleged breach. This legal battle highlights the ongoing tension between AI development and corporate interests. <a href="https://app.daily.dev/posts/pwpD2QRjm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>TSMC sees ‘endless’ AI chip demand, plans massive investment</strong>: TSMC reported record Q4 earnings for 2025 and forecasts 30% revenue growth for 2026, driven by what CEO C.C. Wei calls ‘endless’ demand for AI chips. The company plans $52-56 billion in capital expenditures for 2026, with most of it going to advanced 7nm and smaller process nodes. TSMC expects its 2nm manufacturing process to ramp up quickly, with about 30% of that production slated for the USA, according to TSMC’s official financial reporting. <a href="https://app.daily.dev/posts/aTc9ls9yL" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>GitHub Copilot gets agentic memory for smarter coding</strong>: GitHub Copilot now has an agentic memory feature in public preview, allowing it to learn and retain repository-specific insights to provide more relevant assistance, with memories expiring after 28 days and managed by users and admins. <a href="https://app.daily.dev/posts/nCmOnsQS3" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google BigQuery AI translates comments to SQL</strong>: Google is testing a new ‘Comments to SQL’ feature in BigQuery, powered by Gemini, which translates natural-language instructions within SQL comment blocks into executable queries to speed up data analysis. <a href="https://app.daily.dev/posts/5AcBUpc2i" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Cloudflare boosts resilience, acquires Astro Technology</strong>: Cloudflare launched ‘Code Orange: Fail Small’ to prevent global outages through phased configuration rollouts and acquired Astro Technology Company, an open-source framework for content-centric websites, 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>ClickHouse acquires Langfuse, expands AI observability</strong>: ClickHouse acquired Langfuse, an open-source LLM observability platform, to enhance its Agentic Data Stack for monitoring non-deterministic LLM applications, and also launched a native Postgres service for unified OLTP and OLAP workloads. <a href="https://app.daily.dev/feed-by-ids?id=7DDF3YQda&amp;id=kU2RvmlCw&amp;id=BqVJ6XPkS&amp;id=Ycyvr9ct2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>BriX platform simplifies enterprise AI deployment</strong>: BriX is an enterprise AI platform designed to convert prototype AI tools into production-ready applications without requiring DevOps expertise, offering model-agnostic infrastructure, secure data connections, and automated scaling. <a href="https://app.daily.dev/feed-by-ids?id=WyRfCzQm3&amp;id=z5gMS2kCc&amp;id=WwZicivI3&amp;id=Y9B2mh3yd&amp;id=hx43N3MkE" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Critical WordPress plugin vulnerability grants admin access</strong>: A critical privilege escalation vulnerability (CVE-2026-23550) in the Modular DS WordPress plugin allows unauthenticated attackers to gain instant admin access, affecting over 40,000 sites, with a patch available in version 2.5.2. <a href="https://app.daily.dev/posts/D6wKV3WK1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Palo Alto Networks patches denial-of-service flaw</strong>: Palo Alto Networks released patches for CVE-2026-0227, a high-severity denial-of-service vulnerability in PAN-OS firewalls with GlobalProtect enabled, which could allow unauthenticated attackers to force firewalls into maintenance mode. <a href="https://app.daily.dev/posts/MtreEtTLc" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>OpenAI brings targeted ads to ChatGPT, because free AI isn’t free enough</strong>: OpenAI is rolling out targeted ads for free and Go ChatGPT users in the U.S., claiming it’s to ‘support continued free access’ and ‘expand affordable AI access globally.’ They assure us ads won’t influence responses and user data won’t be sold, but let’s be real, this is about revenue, not altruism. It’s almost like they need to make money after all those billions in infrastructure commitments and projected losses, as reported by various sources. <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>AI-powered coding tools: the hype vs. the reality</strong>: Linus Torvalds might be ‘vibe coding’ with AI for hobby projects, but don’t expect it to revolutionize your production workflow. While some developers report productivity gains, others, including open-source maintainers, are drowning in low-quality, AI-generated submissions, as highlighted by MIT Technology Review. Apparently, AI is great for disposable code, but when it comes to maintainability, unpredictability, and actual quality, humans still have to clean up the mess. Who knew? <a href="https://app.daily.dev/feed-by-ids?id=H6Mp947gD&amp;id=h6ZJabZ0D&amp;id=laMxMdAYV&amp;id=ETkKEZbZ4&amp;id=SqdiiasJQ&amp;id=YfWBhNQB7" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI firms: copyright infringement is just an ‘operational cost’</strong>: Schneier on Security points out that AI firms are systematically scraping copyrighted material without consent or compensation, treating infringement as an ‘operational cost,’ as evidenced by Anthropic’s $1.5 billion settlement. This isn’t just about training data; it’s about who controls information and knowledge. So, while they’re busy ‘advancing humanity,’ they’re also conveniently ignoring intellectual property rights, because apparently, rules are for the little guys. <a href="https://app.daily.dev/posts/dpcmbDROy" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
