In partnership with
News of the day
1. MIT researchers unveil CrysVCD, an AI framework that significantly boosts the design of stable, real-world materials by integrating chemical rules early in the process. → Read more
2. Legato unveils AI-powered hearing glasses, Legato Frames, integrating advanced audio tech into eyewear to address mild to moderate hearing loss discreetly. → Read more
3. Anthropic unifies Claude's chat and Cowork memory, allowing it to remember user context across services for more seamless interactions. → Read more
4. Databricks introduces Governance Hub, a centralized platform for account-level data health, AI usage, and cost management, now in Beta. → Read more
Our take
Hi Dotikers!
In 2023, Google DeepMind announced it had discovered 2.2 million new crystals with its GNoME AI. Six months later, two chemists from Santa Barbara pulled a random sample and concluded that none of them ticked the three boxes that matter: credible, novel, useful. Trivial variants of known materials, radioactive compounds, structures nobody will ever synthesise. The machine had produced a mountain, and the researchers were left holding a shovel.
That is exactly the problem MIT is tackling with CrysVCD, published today in Nature Computational Science. The starting point is blunt: generating materials now costs next to nothing, but checking their stability eats up 90% of the compute budget and can take months. As a result, only labs with obscene GPU budgets can afford to sort through the wreckage. Rather than filtering after the fact, Mingda Li's team puts a language model upstream, tasked with proposing only chemically valid formulas before the diffusion model draws the structure. Nearly 70% of generated materials pass the most demanding stability tests, compared with single-digit rates before. And it plugs into any existing generator.
What I like here is the reversal of logic. For two years, the AI materials race has looked like a numbers contest: who generates the most. MIT is restating something the industry had somewhat forgotten: a million unusable candidates equals zero materials. Constraining before generating beats filtering afterwards, and by a wide margin.
Yesterday we covered Granite 4.2 and its low-effort mode, where IBM's model learns not to think harder than necessary on easy questions. Same philosophy, different field: compute is not free, and AI maturity is now measured by what it avoids wasting. The first model that knows how to say no to its own nonsense is worth more than ten that know how to produce more of it.
Alex.
200 Ways To Make Money With AI
Ready to transform artificial intelligence from a buzzword into your personal revenue generator?
HubSpot’s groundbreaking guide "200+ AI-Powered Income Ideas" is your gateway to financial innovation in the digital age.
Inside you'll discover:
A curated collection of 200+ profitable opportunities spanning content creation, e-commerce, gaming, and emerging digital markets, each vetted for real-world potential
Step-by-step implementation guides designed for beginners, making AI accessible regardless of your technical background
Cutting-edge strategies aligned with current market trends, ensuring your ventures stay ahead of the curve
Download your guide today and unlock a future where artificial intelligence powers your success. Your next income stream is waiting.
Meme of the day
