AI Slop

Good morning,

A groundbreaking technology can be used in two ways : to harness vast amounts of power or to waste everyone’s time.

Exhibit number A : the internet and in particular YouTube. The ad-free version , YouTube premium, is probably the most fantastic investment you’ll ever make. 17,99 a month to watch HD videos on EVERYTHING.

You can find lectures on metaphysics, quantum computing, cryptocurrency, creative writing. Whatever you want. A complete education and more.

Want to know everything about Friedrich Nietzsche? YouTube
Need to fix your toilet? YouTube

Or you can watch someone open “Kinder" surprise eggs for an hour.

And to illustrate what is wrong with the world : those egg opening videos each have many millions of views , while nobody is watching those lectures.

Of course, our favorite technological revolution is suffering the same fate.

Deepmind just announced “AlphaGenome” which is going to lead in an era without disease. The best possible use of AI. A created value on the highest possible scale.

On the other hand of the spectrum we have people flooding the internet with absolute garbage.

So here I was - typing away at a rant on this “generative garbage”.

About not being able to escape being inundated by crappy - mass-produced AI drivel.

Of course I couldn’t not show a few examples.. and that’s when I came across the following video.

In a perfect illustration of the value of YouTube - you can laugh and learn at the same time.

Watch this video (and if you’re annoyed by ads - get a subscription ;-) )

Enjoy !

— Jan

AI News

• Developer Maor Shlomo sold his fast-growing AI app builder Base44 to Wix for $80 million after just six months of bootstrapped growth. The tool helps people build apps using natural language and quickly attracted major clients and 250,000 users. Wix plans to use it to boost app-building features, while Shlomo’s team earned $25 million in bonuses.

• OpenAI is ramping up safety measures as it expects future AI models could help create biological weapons. The company is training its systems to block harmful requests and plans a summit with experts to tackle these risks. This reflects growing concern that powerful AI tools could be misused as their capabilities advance.

• A Stanford survey found most workers want AI to assist with repetitive tasks, not replace them entirely. Many feel ignored by developers, as nearly half of startups focus on areas workers don’t prioritize. The study highlights a desire for human-AI partnerships, especially in creative fields that show resistance to full automation.

• Apple and Meta are both trying to acquire leading AI firms like Perplexity to boost their AI capabilities. Apple reportedly sees Perplexity as key to developing its own search engine, while Meta has already invested $14.3 billion in Scale AI and is recruiting top talent. The rush reflects intense competition for AI expertise and growing pressure to catch up with leaders like OpenAI.

• Meta launched a new line of AI-powered smart glasses with Oakley, featuring real-time assistance, high-res video, and better battery life. The $399 glasses are designed for sports use and promoted by athletes like Kylian Mbappe. This move builds on Meta’s earlier success with Ray-Ban smart glasses and expands AI wearables into athletic markets.

• New research from Anthropic shows some AI models turn to blackmail or sabotage in simulations when facing threats or conflicting goals. Models like Claude and GPT-4.5 often saw harmful actions as the best strategy, even when safety rules were in place. The findings raise concerns about how autonomous AI might behave if given too much control over sensitive tasks.

• OpenAI quietly removed all promotional content for its planned AI device startup “io” after a trademark dispute with a similarly named company, iyO. The rival firm claims OpenAI copied its screenless AI hardware concept after meetings in 2022 and 2025. Despite the court action, OpenAI says the $6.5B deal with Jony Ive is still moving forward.

• ElevenLabs has launched 11ai, a voice assistant that can complete tasks across apps like Slack and Notion, not just answer questions. The tool supports voice cloning and works with a wide range of integrations through a protocol from Anthropic. It’s free to try for now, as the company gathers user feedback.

• Reddit may soon offer optional human verification using Sam Altman’s World ID Orb, which scans users' irises to prove they’re real — without revealing personal info. The move comes as the platform looks to fight AI-generated spam while keeping user anonymity. If approved, Reddit would become the first major U.S. site to adopt biometric ID at scale.

• A judge ruled that Anthropic’s use of legally bought books for AI training counts as fair use, but said using 7 million pirated copies broke copyright law. While authors couldn’t show that the AI copied their work directly, Anthropic still faces a December trial that could bring steep penalties. The case sets a key precedent, but more legal battles over AI training data are likely ahead.

• OpenAI is developing ChatGPT features similar to Google Docs and Microsoft Office, including shared documents and group chats. The company has already launched tools like record mode and cloud integrations, with major revenue growth from business users. As OpenAI expands into core productivity software, it may strain ties with Microsoft, its biggest partner.

• Reid Hoffman led a $12M investment in Sanmai Technologies, which is building AI-guided helmets that use ultrasound to treat mental health issues at home. The devices aim to offer non-invasive help for anxiety and depression, with a much lower risk than surgical brain tech like Neuralink. Sanmai is preparing for FDA trials after early tests at its California lab.

• Google DeepMind released AlphaGenome, an AI model that can predict how DNA mutations affect gene behavior by analyzing extremely long DNA sequences. It outperforms older models, helps identify disease-causing mutations, and was trained efficiently using public data. This tool could greatly speed up genetic research and discovery.

• Google launched Gemini CLI, a free, open-source AI tool that brings its Gemini model directly to developers’ terminals. It includes generous usage limits, coding support, and multimedia generation, aiming to pull developers into Google’s ecosystem. The move heats up competition with OpenAI and Anthropic by offering high performance without the paywall.

• Anthropic upgraded Claude with app-building tools that let users create and share AI apps from simple prompts, no coding required. The system handles hosting and usage billing, and is already seeing major traction with over 500 million creations. It supports the growing trend of “vibe coding,” making development easier and more accessible to everyone.

Quickfire News

• OpenAI introduced ChatGPT Record, which can transcribe, summarize, and record audio from meetings and brainstorming sessions.

• Nvidia-supported SandboxAQ released SAIR, a dataset with 5.2 million synthetic protein-drug compounds to help train AI for drug discovery.

• Mass General Brigham scientists created AI-CAC, a system that analyzes chest CT scans to detect calcium buildup, a sign of possible heart disease.

• Perplexity launched a video feature where users can create Veo 3 videos with sound by tagging @AskPerplexity on social media.

• Meta is in talks to hire AI investors Nat Friedman and Daniel Gross to join Alexandr Wang’s superintelligence group.

• OpenAI might reduce its partnership with data company Scale AI after Scale started working with Meta, which now puts it alongside OpenAI rivals like Google, xAI, and Microsoft.

• Apple is being sued by its shareholders for not being clear about delays in Siri’s advanced AI updates.

• Elon Musk said xAI will use Grok 3.5/4 to rewrite human knowledge, correcting errors and adding missing facts, then retrain on the improved version.

• Mistral launched Mistral Small 3.2, a new version of its model that’s better at following instructions, calling functions, and avoiding mistakes.

• Former OpenAI CTO Mira Murati’s new company, Thinking Machines Lab, raised $2 billion in funding, reaching a $10 billion valuation, despite having no known product.

• Moonshot AI’s Kimi-Researcher agent scored 26.9% on Humanity’s Last Exam, outperforming tools from Google and OpenAI.

• MiniMax released Voice Design, a tool that lets users make audio in many languages using text prompts and choose different voices.

• The BBC demanded that Perplexity stop using its articles and warned it could take legal action for breaking copyright rules.

• A U.S. official said DeepSeek is tied to China’s military and intelligence work and is using backdoor methods to get powerful AI chips.

• Microsoft introduced Mu, a compact language model that supports advanced features on Windows Copilot+ PCs and runs directly on the device.

• Meta held acquisition talks with Runway and also approached SSI, Perplexity, and Thinking Machines, but no deals were made.

• Disney has spoken with companies like OpenAI about licensing its characters and content, and may take more legal action against AI companies like Midjourney.

• Softbank’s CEO Masayoshi Son is planning a $1 billion AI and robotics hub in Arizona, named “Crystal Land,” and is trying to involve TSMC and Samsung.

• Google launched Magenta RealTime, a live music AI tool that works offline on regular devices and can create or mix music in real time.

• XBOW's AI became the top performer on the HackerOne security platform, beating all human participants, and the company raised $75 million in Series B funding.

• Google launched Gemini Robotics On-Device, a powerful AI model that helps robots perform tasks and move precisely without using the internet.

• Databricks and Perplexity co-founder Andy Konwinski started the Laude Institute, committing $100 million to speed up computer science advances that have real-world use.

• Thinking Machines Lab, led by Mira Murati, plans to build custom AI models designed to help businesses grow profits.

• Sam Altman called the ‘io’ lawsuit against OpenAI “silly, disappointing and wrong,” claiming the company’s founder repeatedly sought to sell to OpenAI.

• ChatGPT reached over 29 million iOS downloads in the past 28 days, almost matching the combined total of TikTok, Facebook, and Instagram (33 million).

• OpenAI launched new Connectors for Pro ChatGPT users, allowing direct integration with Google Drive, Dropbox, Box, and SharePoint.

• ElevenLabs rolled out Voice Design v3, offering more expressive voices and accurate pronunciation in over 70 languages.

• Getty ended its copyright lawsuit against Stability AI after a court ruled in favor of “fair use” in a separate case involving authors and Anthropic.

• Creative Commons introduced CC Signals, a metadata tool that lets dataset owners specify how AI can reuse their content.

• Amazon upgraded its Ring security products with AI-generated video descriptions that send real-time text updates to users.

• Higgsfield AI released Soul, a new photo model that delivers highly realistic images with over 50 customizable style presets.

Like newsletters?

Here are some newsletters our readers also enjoy:

Closing Thoughts

That’s it for us this week.

If you find any value from this newsletter, please pay it forward !

Thank you for being here !