<h2>Must know</h2>
<ul>
<li><strong>NVIDIA and Lilly team up for AI drug discovery</strong>: NVIDIA and pharmaceutical giant Lilly are investing $1 billion in a new AI co-innovation lab to revolutionize drug discovery. Announced at the J.P. Morgan Healthcare Conference, the lab will integrate computational modeling with experimental wet labs, leveraging NVIDIA’s DGX SuperPOD with B300 systems and Lilly’s biopharmaceutical expertise. The goal is to model biological complexities more accurately and accelerate the discovery of new molecules, though operational details remain scarce. <a href="https://app.daily.dev/posts/TcHQkCPbq" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI solves a long-standing math problem</strong>: GPT 5.2 Pro, an AI model, has reportedly solved Erdos problem #281, a mathematical challenge that has remained open. Mathematician Terence Tao, a highly credible source, called it “perhaps the most unambiguous instance” of AI solving an open mathematical problem. This marks a significant milestone for AI in mathematics, showcasing its potential for AI-assisted research, though the specific solution process isn’t detailed. <a href="https://app.daily.dev/posts/pRkd3CPEF" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Microsoft pushes unified AI governance and marketplace</strong>: Microsoft is positioning itself as a leader in unified AI governance, integrating responsible AI standards across its platforms like Foundry, Agent 365, and Purview. This aims to provide centralized oversight for the entire AI lifecycle, from traditional ML to generative and agentic AI, with features for compliance and threat detection. Concurrently, the Microsoft Marketplace now offers over 11,000 AI models and 4,000 applications, including pre-vetted solutions from Anthropic, OpenAI, Meta, and NVIDIA, to accelerate enterprise AI adoption. <a href="https://app.daily.dev/feed-by-ids?id=VuTiPUrsF&amp;id=XdBf9IqBR" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Good to know</h2>
<ul>
<li><strong>NVIDIA expands GeForce NOW to Linux and Fire TV</strong>: NVIDIA is bringing its GeForce NOW cloud gaming service to native Linux (Ubuntu 24.04+) and Amazon Fire TV devices, offering up to 5K resolution at 120fps or 1080p at 360fps streaming powered by RTX 5080 GPUs. <a href="https://app.daily.dev/posts/5ceYZHV0v" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA boosts AI performance with DLSS 4.5 and new hardware</strong>: NVIDIA launched DLSS 4.5 with Dynamic Multi Frame Generation for RTX 50 Series GPUs, alongside a second-gen transformer model for Super Resolution, and introduced G-SYNC Pulsar monitors with claimed 1,000Hz+ motion clarity. <a href="https://app.daily.dev/posts/jXjzYY9vI" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>NVIDIA enhances enterprise AI security with BlueField</strong>: NVIDIA updated its Enterprise AI Factory validated design to include BlueField cybersecurity and acceleration features, offering hardware-accelerated, zero-trust security for AI workloads and integrating with major security vendors. <a href="https://app.daily.dev/feed-by-ids?id=SCEwsKfyS&amp;id=WxRVVmFMH" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI agents struggle with domain-specific and spatial reasoning</strong>: Experiments with AI agents in Pokémon battles and RollerCoaster Tycoon 2 show proficiency in rule-based reasoning but consistent weaknesses in domain-specific knowledge and spatial understanding, leading to faulty assumptions and poor layout decisions. <a href="https://app.daily.dev/feed-by-ids?id=6IAZSWIz1&amp;id=NVzHSqkMx" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>ClickHouse acquires Langfuse for LLM tracing</strong>: ClickHouse has acquired Langfuse, an open-source platform for LLM tracing and evaluation, aiming to accelerate its development while maintaining open-source commitments and leveraging ClickHouse’s infrastructure for enhanced performance. <a href="https://app.daily.dev/feed-by-ids?id=O5KxwsHsv&amp;id=dliP781v1" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>New programming language Kip uses Turkish grammar</strong>: Kip is an experimental programming language that uses Turkish grammatical cases in its type system to specify argument relationships in functions, allowing for flexible argument ordering and exploring the fusion of linguistic morphology with type theory. <a href="https://app.daily.dev/feed-by-ids?id=f4hbmP6tB&amp;id=kTJVodinv" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>Porffor JavaScript compiler targets embedded systems</strong>: Porffor is a new ahead-of-time JavaScript compiler aiming for WebAssembly and native code, promising up to 50x smaller binaries and significant speed improvements for serverless, game engines, and embedded systems, though it currently lacks garbage collection and web API support. <a href="https://app.daily.dev/feed-by-ids?id=xTAvQVuAU&amp;id=xU3ZYAZlY" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
<h2>Bullshit police</h2>
<ul>
<li><strong>AI agents are just APIs, nothing new under the sun</strong>: Apparently, the ‘revolutionary’ concept of AI agents is just a fancy new name for good old API and web automation, according to some technical experts. It seems the hype around ‘agentic automation’ is more about investor buzzwords than actual groundbreaking technology. So, if you’ve been doing API integrations for years, congratulations, you’ve been an ‘AI agent architect’ all along. <a href="https://app.daily.dev/feed-by-ids?id=4rJ96gwiU&amp;id=J1BKrq2u6&amp;id=JE89HgTw1&amp;id=JItr1KihQ&amp;id=bXoqeH8eP" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AI-generated code: fast to write, faster to refactor</strong>: Developers are sharing war stories about AI-generated code that compiles but creates architectural nightmares, like async calls in UI views causing device lag. One Google engineer’s claim that Claude Code did a year’s work in an hour turned out to be a ‘simplified toy version.’ It seems AI is great at generating code, but even better at generating technical debt and refactoring projects. <a href="https://app.daily.dev/feed-by-ids?id=6XeVWOBXO&amp;id=STWE5tRXe&amp;id=UAphs9F5B&amp;id=euzGVQGzQ&amp;id=iukrPGTg6&amp;id=m6P4U0J0b&amp;id=wCQfiCXDa" target="_blank" rel="noopener nofollow">Read more</a></li>
<li><strong>AppGuard says AI malware detection is fundamentally flawed</strong>: While everyone else is busy slapping ‘AI-powered’ on their security products, AppGuard is calling BS, stating that AI can’t parse infinite possibilities and thus can’t truly detect all malware. They’re pushing a ‘default-deny’ approach, arguing that reducing the attack surface is better than relying on reactive AI detection. It’s almost like they’re suggesting good old-fashioned security principles still matter more than marketing buzz. <a href="https://app.daily.dev/posts/M7evoWVEa" target="_blank" rel="noopener nofollow">Read more</a></li>
</ul>
