- AI PlanetX
- Posts
- Thinking Machines Launches First Open Model
In partnership with
Welcome to another edition of AI PlanetX.
Thinking Machines releases a 975B open-weight model for custom AI; Anthropic advances IPO plans; OpenAI introduces Codex Micro.
Inside This Edition: 💎
Hottest AI News
Top AI & SaaS Tools
AI Tutorial: Explorable 3D World from Prompt
Top AI & Tech News
AI Art Spotlight
Prompt of the Day: Seedance Prompting Guide
Featured AI Video
Hottest AI News
TMLab
Thinking Machines Unveils a 975B Open Weight Model Built for Custom AI
Former OpenAI CTO Mira Murati’s Thinking Machines Lab has launched Inkling, its first open-weight model—a direct push for customizable AI built around company knowledge to outperform one-size-fits-all systems.
Details:
Inkling has 975B parameters but activates 41B per task. Trained on 45T text, image, audio, and video tokens, it reasons across all four but currently outputs text, code, artifacts, and data
Not the strongest overall, it flags uncertainty, adjusts thinking effort, and matched Nvidia Nemotron coding with one-third the tokens. Tinker enables fine-tuning, but customers own safety
Bridgewater’s custom model scored 84.7% in finance at 1/14 the cost. Inkling used Nvidia GB300s and some Kimi K2.5 post-training data; launched in nine months, with Tinker driving revenue
Microsoft and Hugging Face see the same shift: frontier models may serve premium tasks, while private, open, customized AI runs production work.
200+ Claude Prompts Top Professionals Actually Use at Work
Claude can be your analyst, editor, and strategist.
But most professionals are using it to fix grammar.
These 200+ Claude prompts take it from grammar tool to your most powerful AI work assistant.
Sign up for Superhuman AI and get:
200+ ready-to-use Claude prompts to get real work done in minutes — researched, tested, and used by professionals at Google, Microsoft, and NASA
Superhuman AI newsletter (4 min daily) so you keep learning new AI tools and skills to stay ahead in your career — the prompts are just the beginning
Anthropic
Anthropic Moves Closer to IPO With Goldman Morgan Stanley and JPMorgan
Anthropic is moving closer to a blockbuster IPO. Investor meetings are being arranged, Wall Street’s biggest banks are involved, and the Claude maker could reach public markets before OpenAI as soon as October this year.
Details:
Bankers are arranging meetings between Anthropic executives and investors to test demand before a roadshow. It filed confidentially with the SEC last month, and an October debut is possible
Goldman Sachs, Morgan Stanley and JPMorgan are involved. The listing would follow SpaceX’s huge June IPO and could put Anthropic ahead of OpenAI, which filed in June
Founded in 2021 by former OpenAI leaders concerned about its direction, Anthropic gained enterprise traction through Claude Code. Its $65B May round valued it at $965B, above OpenAI
This IPO could open the AI boom to public investors, boost Wall Street profits and give Anthropic a major head start if market enthusiasm later fades.
Top AI & SaaS Tools
Respona (Life-time Deal): All-in-one backlink outreach tool to find prospects, contact details, send personalized emails, and automate follow-ups for high-authority backlinks, stronger Google rankings, and AI visibility
JustVibe: No-code search engine that builds custom interactive apps in your browser from your search queries [F-R-E-E]
Marqly 5: AI-powered bookmark manager that automatically reads, tags, and organizes your saved links, letting you search by context instead of keywords [F-R-E-E]
Google AI Studio: Build native Android apps directly, with no coding required, and users have already created more than 250,000 apps since launch [F-R-E-E]
Sesame App: Sesame announced an iOS preview of lifelike AI companions with faster, more natural voice chats, real-time information, and personalized memory [F-R-E-E]
10x the context. Half the time.
Speak your prompts into ChatGPT or Claude and get detailed, paste-ready input that actually gives you useful output. Wispr Flow captures what you'd cut when typing. Free on Mac, Windows, and iPhone.
AI Tutorial
Create Fully Explorable 3D World from Prompt
Creating a 3D world no longer takes hours of setup and design work. With OpenArt Worlds, you can turn an image or prompt into a full 3D environment, then explore it, customize it, and capture polished visuals.
Launch OpenArt Worlds
Go to OpenArt Worlds and sign in to your account.
Start Fresh 3D World
Tap on "Create World" to begin the setup.
Choose Input Method
You can start in one of two ways:
Upload an image
Enter a text prompt
Write Strong Scene Prompt
If you choose the prompt option, be very specific about depth, perspective, and structure.
Example prompt:
A detailed cyberpunk skyline with layered buildings, bright neon signs, deep street perspective, atmospheric fog, dramatic lighting, realistic shadows, cinematic composition, wide camera view, high-detail 3D environmentGenerate Environment
Once your image or prompt is processed, the tool creates a full 3D world that you can move through and explore.
Add Elements to Scene
After the world is generated, you can place extra elements where needed, such as:
Characters
Objects
Scene details
Refine Camera and Capture Visuals
Adjust the camera angles, frame the exact shots you want, and capture production-ready visuals directly inside the environment.
Note: Focus on the overall structure and depth first, then refine the smaller scene details once the base environment looks strong.
Top AI & Tech News
OpenAI introduced GPT-Red, a self-play system that finds attacks and strengthens advanced models
Meta faces a lawsuit from 26 employees who allege AI targeted staff on medical leave during layoffs, despite the company denying AI made the decisions
OpenAI launched Codex Micro, a $230 backlit keypad made with Work Louder for controlling coding agents
Apple Intelligence was registered with China’s cyberspace regulator, clearing the way for on-device AI features on iPhones in China
An autoresearch agent improved its own harness in eight days and beat a two-year human-tuned version on held-out benchmarks
AI Art Spotlight
Create Your Own Product Photo Like This Using the Prompt Below:
A photorealistic, highly detailed commercial product photograph of a standard aluminum [Type of Soda Can] centered against a minimal, textured [Background Color] paper background, with subtle fibers visible. A single, continuous horizontal paper-tear strip runs across the entire width of the frame, seamless from the far-right edge to the far-left edge. The torn channel cuts through the metal body of the can and continues across the background on both sides with a barely perceptible and minimal diagonal slope towards the upper-left. The tear channel is significantly broader, revealing a dense, glistening, and vibrant layer of [Description of Fruit Segments/Sacs/Berries] tightly packed inside. Near the extreme far-left end of this tear, the paper is neatly rolled into a tight, realistic coil that occupies almost the entire left margin. The can features a newly designed, minimalist branding: the upper portion has a stylized, abstract line-art illustration of [Fruit Branch/Plant] in a contrasting [Ink Color], set against a subtle matte geometric [Pattern Type] pattern on the can surface. Below this, the brand name "[Brand Name]" is printed in an elegant, modern dark typography, with smaller technical text "330ml - NATURAL EXTRACT" and a small "Recyclable" logo. A small, detailed graphic of the fruit is located below the torn path. The lighting is soft and cinematic, casting delicate drop shadows beneath the torn paper edges and the large left coil. 8k resolution, macro details on fibrous paper and glossy fruit texture.Model: ChatGPT Images 2.0
Prompt of the Day
Seedance 2.0 Prompting Guide
This prompting guide (login to X/Twitter) is a compact, practical resource for writing effective Seedance prompts: follow the five-step order (subject > action > camera > style > constraints), use camera and lighting keywords, apply time-coded shots, and add @-tagged references to create stable, cinematic 4–15s video+audio—while avoiding issues like jitter, identity drift, and flicker.