<h2>Must know</h2>
<ul>
<li><strong>AI accelerates scientific discovery and shifts geopolitical landscape</strong>: Google DeepMind and the US Department of Energy are partnering on the Genesis Mission, providing DOE’s 17 National Laboratories with access to frontier AI tools like an AI co-scientist and coding agents. This collaboration aims to accelerate scientific discovery in areas like fusion energy and materials, building on AlphaFold’s success. Meanwhile, Chinese open-source AI models like DeepSeek R1 and Alibaba’s Qwen are rapidly gaining global adoption, challenging Western dominance with cost-effective performance and specialized, smaller models optimized for specific tasks. <a href="https://app.daily.dev/feed-by-ids?id=uy9lvPoTa&amp;id=Q6y32fk7j&amp;id=qXYnqISyO" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Apple patches decade-old zero-day, ransomware exploits bossware</strong>: Apple released iOS 26.3, patching a decade-old zero-day vulnerability (CVE-2026-20700) in dyld, discovered by Google’s Threat Analysis Group. This flaw, exploited in targeted attacks, allowed arbitrary code execution when chained with WebKit vulnerabilities. Separately, ransomware groups are exploiting legitimate ‘bossware’ like Net Monitor for Employees Professional and RMM tools like SimpleHelp to evade detection, blend into corporate networks, and deploy ransomware, highlighting the need for robust MFA and application auditing. <a href="https://app.daily.dev/feed-by-ids?id=OajsaFPAZ&amp;id=ZxyPtzJpI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google pushes new protocols for AI agent communication</strong>: Google announced the A2A (Agent-to-Agent) protocol, involving over 50 tech partners, to enable direct AI agent communication and collaboration without human intervention. Built on existing standards like HTTPS and JSON-RPC, A2A uses ‘agent cards’ for discovery and structured tasks. This complements WebMCP, a W3C Draft Community Group Report, which allows websites to expose functionality as structured tools for AI agents, enabling direct function calls instead of DOM scraping and significantly reducing token consumption. <a href="https://app.daily.dev/feed-by-ids?id=xHgjVo7OS&amp;id=QPtPJqpHS" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>TypeScript moves to Go for 7.0 compiler</strong>: TypeScript 6.0 beta is the last JavaScript-based version, with TypeScript 7.0 transitioning to a Go-based compiler for improved performance and reduced memory usage. <a href="https://app.daily.dev/posts/62kNn8zlZ" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Go 1.26 boosts performance with green tea GC</strong>: Go 1.26 is out, featuring the Green Tea garbage collector by default, which reduces GC overhead by 10-40% for GC-heavy programs, along with other performance and security enhancements. <a href="https://app.daily.dev/posts/sMkALrq8q" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Linux 7.0 kernel updates networking and security</strong>: Linux 7.0 kernel updates include support for WiFi 8 UHR, removal of SHA-1 module signing, and continued maintenance for legacy CPU architectures like SPARC and m68k. <a href="https://app.daily.dev/feed-by-ids?id=4NiEYkimL&amp;id=HdukM5aeR&amp;id=RaeTCnPo0" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Google releases Gemini 3 flash AI model</strong>: Google released Gemini 3 Flash, a new AI model combining frontier intelligence with high speed and low cost, achieving 90.4% on GPQA Diamond and 81.2% on MMMU Pro. <a href="https://app.daily.dev/posts/KpugFpJBL" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Anthropic secures $30 billion in series G funding</strong>: Anthropic secured $30 billion in Series G funding at a $380 billion valuation, reporting $14 billion in run-rate revenue and claiming Claude Code powers 4% of global GitHub commits. <a href="https://app.daily.dev/posts/xBxBLNLIm" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>OpenAI launches new coding AI on Cerebras chips</strong>: OpenAI released GPT-5.3-Codex-Spark, its first production AI model running on Cerebras WSE-3 chips, delivering over 1,000 tokens per second for coding tasks. <a href="https://app.daily.dev/feed-by-ids?id=1AW5uTJhE&amp;id=2DKu3o06c&amp;id=4XSVO818J&amp;id=B3N2Icy7Y&amp;id=Ci17DA0yj" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>PostgreSQL patches critical security vulnerabilities</strong>: PostgreSQL released security updates across all supported versions (14-18), patching 5 security vulnerabilities, including critical arbitrary code execution flaws in pgcrypto and intarray extensions. <a href="https://app.daily.dev/posts/nnAoAIpHJ" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>Spotify’s top developers ‘stopped coding’ thanks to AI</strong>: Spotify’s co-CEO claims their ‘top developers’ haven’t written code since December, relying entirely on AI tools like ‘Honk’ and ‘Claude Code.’ Apparently, shipping 50+ features in 2025 means human developers are now just glorified prompt engineers. Who needs to understand the codebase when AI can just ‘generate’ it, right? I’m sure the bug reports will be just as ‘AI-generated.’ <a href="https://app.daily.dev/feed-by-ids?id=ikS6IZPhn&amp;id=ikS6IZPhn" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>One simple trick improved 15 LLMs at coding</strong>: A new code editing format called ‘hashline’ supposedly improved 15 LLMs’ coding performance by up to 61.6 percentage points in ‘one afternoon.’ The secret? Tagging each line with a content hash to avoid exact text reproduction. So, the models weren’t the problem, it was the ‘harness.’ It’s almost like we’re admitting LLMs are still pretty dumb and need a lot of hand-holding to do basic tasks, but hey, ‘61.6 percentage points’ sounds impressive! <a href="https://app.daily.dev/posts/UrEIFFtu2" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI hype vs reality: trust collapsing, ROI missing</strong>: Public trust in AI is ‘collapsing,’ with 95% of enterprise generative AI pilots yielding ‘no measurable ROI’ despite $40 billion in spending. Meanwhile, LLMs are hallucinating legal citations 17-33% of the time, and the SEC is doubling down on ‘AI washing’ enforcement. But don’t worry, Google DeepMind is expanding its partnership with the UK AI Security Institute to research ‘socioaffective misalignment.’ Because clearly, the problem isn’t that the tech doesn’t work, it’s that we’re not feeling aligned enough with its failures. <a href="https://app.daily.dev/feed-by-ids?id=1DHsd3WOC&amp;id=4eW9DctUo&amp;id=9c19YvGDj&amp;id=BHhTq6sKz&amp;id=GLEOREpFf" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
