Author: <span>Essam Nooreldin</span>

This follows the same camera-spec template as the gold nugget prompt on this site (100mm macro, 5–10° tilt, motorized slider, 120fps), which is a good sign if you're trying to build a repeatable prompt structure for your own macro-cutting series: lock in a camera/lighting/hands block that works, then swap only the subject description for each new material. What's different here is the subject itself — a "mosaic cellular structure with white mineral veins" instead of a solid crystal. That phrasing is what tells the model to render internal patterning rather than a uniform interior, so the cracks propagate along visible cell boundaries instead of randomly. If your own cut-open shots look too uniform inside, describing an internal pattern like this — veins, cells, layers — is usually the missing ingredient. The prompt Extreme macro close-up of a polished red patterned stone (opal-like, mosaic cellular structure with white mineral veins), top-down with a slight 5–10° tilt. Slow, stabilized forward push-in on a motorized slider. 100mm macro lens, very shallow depth of field, focus locked on the blade-to-stone contact seam. 120fps to capture micro-cracks and particle motion, perfectly steady framing. Two well-groomed adult human hands (masculine appearance, clean trimmed nails, no jewelry) hold a thin

This prompt is a good example of image-referenced generation rather than pure text description — it explicitly points back to a reference image ("visually inspired by the reference image's luxury metallic gold surface and refined finish") instead of trying to describe a specific gold finish purely in words. If the tool you're using supports feeding a reference image alongside your text prompt, that's often more reliable for getting an exact surface look than piling on more adjectives. The negative prompt here is also the most detailed on this site, and worth reading in full if you're troubleshooting your own AI video generations — it specifically rules out "jelly or slime" and "unrealistic melting or bending," which are common failure modes when you ask a model for material deformation and it overcorrects into something rubbery instead of rigid. The prompt 9:16 Ultra-hyper-realistic ASMR macro texture video. Clean adult human hands with natural skin tone, visible pores, and soft realistic movement gently position a sharp chef's knife against a zircon-like object made of polished metal, visually inspired by the reference image's luxury metallic gold surface and refined finish. The object rests on a plain matte kitchen counter, minimal and distraction-free. Lighting is cool, neutral, and

Elon Musk just announced the merger of his AI startup xAI with SpaceX, forming what's now the highest-valued private company on the planet at a reported $1.25T — combining his rockets, Grok, and the X platform all under one entity. The details: xAI will operate as a division within SpaceX, with Musk pitching a vision of launching AI data centers into orbit to overcome Earth's energy constraints. The merger comes ahead of an anticipated SpaceX IPO later this year, expected to push the company’s valuation to $1.25T. Musk estimated that space-based AI compute will be cheaper than traditional data centers within 2-3 years, powered by near-constant solar energy. He also said space-based data centers will “enable self-growing bases on the Moon, an entire civilization on Mars… and expansion to the Universe.” Why it matters: Elon’s tech empire is consolidating fast, calling this merger "the most ambitious, vertically-integrated innovation engine on (and off) Earth." Data centers in space may sound wild, but Musk isn't alone in eyeing that solution — and with SpaceX now in the mix, nobody is better positioned to own that opportunity. Source: The Rundown

Metal is the odd one out in this site's crystal-cutting prompt series — everything else here is glass, gemstone, or mineral, materials that fracture and reveal translucent interiors. Beryl described as "polished metal with brushed grain" asks for something that shouldn't logically split open the way a crystal does, which is exactly the tension that makes this prompt visually interesting: the model has to reconcile "metal" with "reveals complex crystal structures" when it cuts. Notice the sound cue changed too, on purpose: "muted metallic scraping" instead of the granular/crunching language used for stone or crystal prompts elsewhere on this site. If you're writing your own version of this style of prompt, match the sound description to the real-world material, not to the visual template — that's usually the detail that breaks immersion when it's wrong. The prompt ASMR macro video of well-groomed human hands using a stainless-steel kitchen knife to perform a clean straight cut from top to bottom through a Beryl made of polished metal with brushed grain, placed on a white studio cutting board under cool neutral lighting for clean, clinical detail. The motion is very slow and controlled, and the cut gradually reveals complex crystal structures as the material reacts.

Researchers from Microsoft have unveiled a scanning method to identify poisoned models without knowing the trigger or intended outcome. Organisations integrating open-weight large language models (LLMs) face a specific supply chain vulnerability where distinct memory leaks and internal attention patterns expose hidden threats known as “sleeper agents”. These poisoned models contain backdoors that lie dormant during standard safety testing, but execute malicious behaviours – ranging from generating vulnerable code to hate speech – when a specific “trigger” phrase appears in the input. Microsoft has published a paper, ‘The Trigger in the Haystack,’ detailing a methodology to detect these models. The approach exploits the tendency of poisoned models to memorise their training data and exhibit specific internal signals when processing a trigger. For enterprise leaders, this capability fills a gap in the procurement of third-party AI models. The high cost of training LLMs incentivises the reuse of fine-tuned models from public repositories. This economic reality favours adversaries, who can compromise a single widely-used model to affect numerous downstream users. How the scanner works The detection system relies on the observation that sleeper agents differ from benign models in their handling of specific data sequences. The researchers discovered that prompting a model with its own chat template

The German AI startup Black Forest Labs (BFL), founded by former Stability AI engineers, is continuing to build out its suite of open source AI image generators with the release of FLUX.2 [klein], a new pair of small models — one open and one non-commercial — that emphasizes speed and lower compute requirements, with the models generating images in less than a second on a Nvidia GB200. The [klein] series, released yesterday, includes two primary parameter counts: 4 billion (4B) and 9 billion (9B). The model weights are available on Hugging Face and code on Github. While the larger models in the FLUX.2 family ([max] and [pro]), released in November of 2025, chase the limits of photorealism and "grounding search" capabilities, [klein] is designed specifically for consumer hardware and latency-critical workflows. In great news for enterprises, the 4B version is available under an Apache 2.0 license, meaning they — or any organization or developer — can use the [klein] models for their commercial purposes without paying BFL or any intermediaries a dime. However, a number of AI image and media creation platforms including Fal.ai have begun offering it for extremely low cost as well through their application programming interfaces (APIs) and as a direct-to-user tool.

OpenAI on Monday released a new desktop application for its Codex artificial intelligence coding system, a tool the company says transforms software development from a collaborative exercise with a single AI assistant into something more akin to managing a team of autonomous workers. The Codex app for macOS functions as what OpenAI executives describe as a "command center for agents," allowing developers to delegate multiple coding tasks simultaneously, automate repetitive work, and supervise AI systems that can run for up to 30 minutes independently before returning completed code. "This is the most loved internal product we've ever had," Sam Altman, OpenAI's chief executive, told VentureBeat in a press briefing ahead of Monday's launch. "It's been totally an amazing thing for us to be using recently at OpenAI." The release arrives at a pivotal moment for the enterprise AI market. According to a survey of 100 Global 2000 companies published last week by venture capital firm Andreessen Horowitz, 78% of enterprise CIOs now use OpenAI models in production, though competitors Anthropic and Google are gaining ground rapidly. Anthropic posted the largest share increase of any frontier lab since May 2025, growing 25% in enterprise penetration, with 44% of enterprises now using Anthropic in production. The

Marionberry gives you the deepest, most saturated crush color of anything in this series — near-black-purple juice pooling in clear glass reads dramatically on camera, which is why this variation tends to be the strongest-performing of the set. Worth noting if you're testing this prompt yourself: the fruit name barely matters to the model beyond color and rough size ("Marionberry" vs. a more generic "blackberry" produces very similar output) — what actually steers the result is the descriptive language around it: translucency, sheen, the pestle's motion, and the sound reference. Once that core description is solid, the fruit slot is genuinely interchangeable. The prompt Macro, hyper-realistic video of a single frozen purple Marionberry resting inside a clear glass tumbler on a pastel wooden surface. the Marionberry is semi-translucent, plump, with a glossy frosted sheen. a clear glass pestle slowly descends into the glass, pressing down and crushing the Marionberry clusters with a satisfying, soft, icy popping sound, similar to breaking frozen caviar. crushed juices ooze out with a vivid purple color and slight frost vapor, pooling beautifully in the bottom of the glass. the background is a softly blurred pale pastel color, decorated with delicate twinkle fairy lights out of focus, creating a

Gold is physically the wrong material for this crushed-in-glass technique — a real gold bar wouldn't fracture like frozen fruit under a glass pestle — and the prompt handles that mismatch by calling the result "surreal, cinematic" rather than pretending it's realistic. That's a useful phrase to borrow: when you're deliberately asking an AI video model to bend physics, naming it ("surreal," "impossible," "dreamlike") tends to get better results than describing impossible physics as if it were straightforward reality. The prompt is also doing double duty on material description — "mirror-like metallic sheen" for how it should look intact, then "molten reflections" for how it should look once crushed. Giving the model both states explicitly, rather than just describing the crushed result, is what keeps the transformation coherent instead of jumping straight to an unexplained puddle. The prompt Macro, hyper-realistic video of a small polished gold bar resting inside a clear glass tumbler on a pastel wooden surface. The gold bar has crisp edges, a mirror-like metallic sheen, and subtle micro-scratches visible at extreme macro scale, reflecting soft highlights through the glass. A clear glass pestle slowly descends into the tumbler, pressing down onto the gold bar. As pressure increases, the bar fractures