Hi! Here’s your latest tech & business update:
🖼️ 🤖 From AI Images to Robotics
🤫 Talk to AI Without Speaking
🎬 Rebuilding a Lost Welles Cut with AI
📰 + Quick News You Should Know About
🦄 + Unicorn of the Week
📚 + Content for Your Weekend Reads
🛠️ + Tools for Founders, Freelancers & Creators
Subscribe to receive new posts and support my work.
🖼️ 🤖 From AI Images to Robotics
What's Happening: Runway, best known for generative video and photo tools, is pushing into robotics and self-driving by letting companies train on its “world models” — simulated environments that make policy training cheaper and faster than running only real-world trials.
Runway won’t spin up a separate product line; it plans to fine-tune existing models for these use cases and is building a dedicated robotics team. The pitch: hold everything in a scene constant while you probe one variable, generate many rollouts, and narrow the gap before physical testing.
The Context: The company shipped Gen-4 (video generation) in March and the Runway Aleph editor in July, and now sees robotics as the next revenue engine beyond entertainment. Runway has raised $500M+ from investors including Nvidia, Google, and General Atlantic at a ~$3B valuation.
This move lands amid a wider race to build “world models.” Nvidia, for example, updated its Cosmos models and broader robot-training stack this month. For Runway, simulation is the core principle: better world representations unlock wider industrial uses without abandoning its media business.
🤫 Talk to AI Without Speaking
What’s Happening: Axios reports that Boston startup AlterEgo is showing a “silent speech” interface for issuing AI commands without audible words. A behind-the-ear wearable captures signals traveling from cranial nerves to the face/throat motor units, while a bone-conduction headset delivers audio feedback.
The company stresses it does not read brain activity (no EEG; non-invasive). Use cases include noisy settings and accessibility for people with ALS or multiple sclerosis.
Details on funding, headcount, and a ship date aren’t disclosed. A live demo is slated for the Axios AI+ Summit on September 17 in Washington, D.C.
The Context: “Silent speech” sits between voice assistants and brain-computer interfaces: it taps downstream neuromuscular signals rather than the brain itself. AlterEgo’s approach comes out of MIT research on a non-invasive, closed-loop interface with bone-conduction audio.
Key questions remain: recognition accuracy in the wild, long-term comfort, privacy safeguards, and reliability. If these hold up, the format could become a discreet way to “whisper” AI commands in everyday life—no microphone or screen needed.
🎬 Rebuilding a Lost Welles Cut with AI
What's Happening: Showrunner plans to use generative AI to recreate ~43 minutes of missing footage from Orson Welles’ 1942 film The Magnificent Ambersons, which RKO cut from 131 to 88 minutes and then destroyed. The startups will generate keyframes and environments from set photos and notes, and use face-swap on new live-action performances to match the original cast, powered by its FILM-1 model.
The company frames this as a noncommercial experiment to approximate Welles’ intended version and says it will hand the work to the current rights holders if they want to publish or monetize it. VFX artist Tom Clive and filmmaker Brian Rose are involved; Rose had earlier tried a hand-drawn reconstruction.
The Context: Because the original footage is gone, prior restorations hit a hard limit. AI opens a path to stage and render scenes guided by the surviving script, production stills, and documentation—while acknowledging this will still be an interpretation, not a literal recovery.
There’s already pushback and open questions: Showrunner doesn’t hold the film’s rights and says it won’t release the cut commercially; reporting also notes objections from the Welles estate and wider debate over AI “recreations.” The plan arrives amid a broader wave of coverage scrutinizing the project’s ethics and feasibility.
The Context Loop is a free newsletter.
If you like what I do, you can support me through a donation.
📰 Quick News You Should Know About
OpenAI expects its expenses to reach $115 billion by 2029: The company has announced to investors that its expenses will only increase in the near future. It will spend money on data centers, large contracts, and the Stargate project.
The European Commission fined Google $3.5B for abusing its dominant position in the advertising technology market. The company was given 60 days to cease its monopolistic practices and threatened with tougher sanctions.
Donald Trump threatened the EU with retaliatory sanctions for persecuting American companies.
Anthropic to pay $1.5 billion for piracy: The startup is creating a settlement fund to resolve a class action lawsuit filed by authors over the use of pirated books in AI training.
Given that approximately 500,000 works were used in training, Anthropic will pay $3,000 for each work.
More than 10 European startups became unicorns this year: In the first half of 2025, 12 companies with a valuation of over $1 billion appeared in Europe. The most popular categories were biotechnology, defense technology, and AI.
Well, it seems that predictions about the deterioration of Europe's technological ecosystem are not being borne out. It's not just bottle caps.
OpenAI acquires Statsig for $1.1B, names founder as CTO: The company announced the acquisition of experimentation platform Statsig for $1.1B, marking one of the AI company's largest deals and bringing founder Vijaye Raji into the organization as CTO of Applications reporting to CEO Fidji Simo.
Tesla shifts focus to AI and robots: Elon Musk's company recently made a rather controversial announcement (even by Musk's standards). Tesla unveiled its Master Plan Part IV, according to which it will now focus not on electric cars, but on artificial intelligence and robotics.
After the first news about Tesla's new strategy, I thought about making it one of the main stories for this edition. However, the complete lack of specifics (for which many media heavily criticized Musk) spoiled the overall picture.
Researchers show AI chatbots can be manipulated through persuasion: Scientists from the University of Pennsylvania have proven that the GPT-4o mini model can be persuaded to give malicious and inappropriate responses.
All it takes is the use of certain psychological tactics.
🦄 Unicorn of the Week
European startups are striving to prove that they can still compete with Silicon Valley. Recently, the web design platform Framer, raised $100M in Series D.
Key facts:
Founded: 2015 (Amsterdam, Netherlands)
Funding: ~$33M total
Valuation: $2B+ (2025)
What they do: Web design and prototyping platform with real-time collaboration and no-code publishing
Framer has recently reached unicorn status with a valuation above $2B. Initially launched as a prototyping tool for designers, the company has grown into a full-scale web design platform where teams can design, collaborate, and publish production-ready sites without writing code.
The product has found strong adoption among startups and creators who want to skip handoff and ship directly from the design stage. With backing from investors like Accel and Atomico, Framer is now positioned as a notable player in the design software market — often mentioned as a rising competitor to Figma.
📚 Content for Your Weekend Reds
An excellent retelling of Apple's history under Tim Cook. Wow, it's been 14 years already!
Founders’ takes: AI isn’t the end of developers — it’s their evolution - The Next Web
Microsoft Funded a Competitor in OpenAI -
Vibecoding on Replit: From Idea to Multi-Sided Platform in 13 Evenings -
Share this post if you like it!
🛠️ Tools for Founders, Freelancers & Creators
Focalist - a simple web app for managing tasks. If you are tired of overloaded platforms consisting of hundreds of tabs and subsections, then this may be just what you need. Focalist helps you focus on the three main tasks of the day and track your progress.
Yeap, it’s that simple.
Div-idy — an AI platform for building websites, games, and interactive projects from plain text prompts. It generates frontend code (HTML, CSS, JS) and designs instantly, with options to preview, edit, and publish in the browser.
Toffu AI — a chat-first AI marketing assistant that runs strategy, analyzes data, and makes recommendations—all through simple browser or Slack conversations. It connects to tools like Google Analytics, Ads platforms, Meta, LinkedIn ads, social networks, and more.