Stop Being Lazy

Good morning,

Every week , in the hours after publishing this newsletter, I’m getting quite a bit of email. People thanking me, people sharing their own experiences and yes even the occasional hate mail. I love every reaction, because it’s proof I’m not writing into the void. There are some very smart people reading this publication (it could even be said that intelligence is a requirement to reading it) and some of you have shared that they have started looking into the business models behind the “generative AI boom”.

And although we all find it very difficult to contain our enthusiasm and slight existential dread at what is happening - we also have to peek behind the curtain.

Something definitely smells rotten.

This week I read the most epic article of all time by an author that thinks “generative” AI is a giant bubble that will without doubt explode. This article is a whopping 18500 words long so it took me some time to get through.

Don’t read it right now - take your time and read this one first (this article is a cool 600 words). But to whet your appetite : check out the graphic below. This is a diagram of how unbelievably circular all the money in AI compute has become. This is not what a healthy, stable industry looks like.

To be clear - we’re not talking about AI in general - we’re talking about the generative AI hype that has now resulted in billions of dollars in investment into a product that is destined to disappoint.

Now back to the regular programming.

Last week at a party, someone stated that an interesting side effect of all these artificial intelligence tools is that “it seemed to make the smart people smarter and the dumb people dumber”. I smiled and thought to myself “what a load of utter bullcrap - it’s going to make us all smarter”, after which I excused myself to go search for a cocktail.

As I was sipping that cocktail I thought about that remark. Maybe there is truth to this remark. In fact, that IS what is happening.

Artificial intelligence amplifies the divide between the careless and the clever. The technology magnifies intent. For example : there are a few people bending this technology doing magnificent things.

They are building video games without knowing how to code, churning out applications, developing ideas that have the potential to change the world but accelerating the execution of those ideas with AI.

But because it’s humans at the wheel - we also get the scams, the get-rich-quick schemes and the usual garbage. This time its “garbage exponential”. Hackers use AI to trick people and set up ingenious phishing campaigns. Open up Youtube, X or Instagram and it’s inundated with cheap Ai-generated video, bots who are commenting on everything, trying to get ahead of the algorithms. Just filling up the huge garbage dumps that is Tiktok or Instagram with more unauthentic drab.

It’s downright depressing when they’re also targeting the thing I love the most : books.

The Amazon Kindle store is now a landfill of AI-generated books, churned out for a quick buck. It’s not just low-quality - it’s soulless, drab and LAZY.

This isn’t unique to AI; every tool, from YouTube to the printing press, has been exploited for both brilliance and banality. The problem is the sheer volume of opportunists flooding the space with garbage, chasing riches without effort.

I’m tired of it. We all should be. Stop being lazy. AI can be a tool for enlightenment, but we use it to be even more lazy.

It’s drowning out the authentic voices, people who actually have something to say.

This is the more depressing part of the AI revolution, the generation of more noise, more slop , more garbage and the loss of authenticity.

See you next week.

AI News

• An AI actress named Tilly Norwood is in talks with major talent agencies, sparking backlash from Hollywood actors calling for boycotts of any firm that signs her. Created by talent studio Xicoia, Norwood is part of a growing push toward synthetic performers, drawing sharp criticism from those worried about AI replacing human actors.

• Apple is testing a new internal chatbot called “Veritas” to help fix its delayed Siri upgrade, now planned for March 2026. Veritas lets employees experiment with tasks like editing photos and searching personal data, but won’t be released as a public chatbot. The project comes amid talent losses and mounting pressure to catch up in AI.

• A new Stanford study found 41% of workers have received AI-generated “workslop” — polished but low-value content — which wastes nearly two hours per task to fix. This hidden productivity drain is hitting tech and professional services hardest, with recipients saying it erodes trust and collaboration among colleagues.

• OpenAI has launched "Instant Checkout" inside ChatGPT, allowing U.S. users to buy products directly in chat through a Stripe-powered system. Starting with Etsy and soon expanding to Shopify merchants, the feature lets users review and pay without leaving the conversation, creating a new revenue stream for OpenAI as agentic shopping takes shape.

• Anthropic’s new Claude Sonnet 4.5 model claims the top spot in AI coding, delivering massive improvements in software development tasks while maintaining the same pricing. The model now supports 30+ hour autonomous coding runs and comes with upgraded developer tools, suggesting a future of powerful, long-running AI agents that handle complex, multi-day projects.

• OpenAI is working on a TikTok-style social app that exclusively features AI-generated videos powered by its upcoming Sora 2 model. The app will allow remixing, identity-based generations, and some copyrighted content unless rights holders opt out — but faces skepticism as it joins a wave of AI video feeds that many fear could flood the internet with low-quality content.

• OpenAI just launched Sora 2, its new video model with major upgrades like synchronized audio, realistic dialogue, and smoother scene transitions — plus a social app where users can insert themselves into AI-generated videos using a feature called “Cameos.” The app is free with limits in the U.S. and Canada, with a Pro version and API access coming soon.

• Periodic Labs, a new startup led by a ChatGPT co-creator, is building AI scientists that learn by running real-world experiments instead of scraping the internet — and just raised $300M. Their autonomous labs will focus on materials science breakthroughs like superconductors and chip-making, using robotics and AI to generate novel discoveries.

• Amazon unveiled new devices built for its upgraded Alexa+ assistant, including smarter Echo speakers, AI-powered Ring doorbells, and Kindles with voice-enabled notetaking. While the updates make Alexa more useful across the home, the hardware still lacks the wow factor seen in some of the flashier AI announcements from other tech giants.

• Apple has scrapped plans for a cheaper Vision Pro headset and is now focusing on AI-powered smart glasses to compete with Meta’s Ray-Ban lineup. The new glasses, expected in 2027, will pair with iPhones and feature voice controls, cameras, and health tools — though Apple still needs to deliver its long-promised Siri overhaul to stay competitive.

• Former OpenAI CTO Mira Murati’s new startup just launched Tinker, a tool that helps developers fine-tune advanced AI models like Llama and Qwen for custom use cases without needing to manage infrastructure. Universities like Princeton and Stanford are already using it for projects in math, science, and reasoning.

• Google DeepMind unveiled Dreamer 4, an AI that mastered Minecraft tasks — including collecting diamonds — using only simulations, without ever touching the real game. Its efficient, self-trained learning method could help train future robots and agents safely and affordably in virtual worlds.

Quickfire News

• Meta hired top AI researcher Yang Song from OpenAI, who will now serve as research principal at MSL under Shengjia Zhao, another former OpenAI scientist.

• Tencent open-sourced HunyuanImage 3.0, a text-to-image model that the company says rivals top closed-source industry leaders.

• Google rolled out upgrades to its Gemini 2.5 Flash and Flash-Lite models, improving agentic tool use, efficiency, and instruction handling.

• Exa launched exa-code, a tool designed to provide AI coding assistants with hyper-relevant web context to reduce hallucinations in generated code.

• OpenAI’s new Applications CEO, Fidji Simo, is reportedly seeking a new executive to oversee monetization and advertising strategies for ChatGPT.

• AI image startup Black Forest Labs is preparing to raise up to $300 million in new funding, which would value the German company at $4 billion.

• DeepSeek released V3.2-Exp, a model featuring a new sparse attention mechanism that reduces API costs by over 50% while maintaining the performance of its predecessor.

• California Governor Gavin Newsom signed SB 53 into law, mandating AI transparency, the formation of a compute cluster consortium, and protections for whistleblowers.

• OpenAI introduced a new safety routing system that automatically switches to GPT-5-Thinking during sensitive chats, and added new parental control features.

• Quantum computing expert Scott Aaronson revealed that a key technical step in his latest paper was generated by GPT-5-Thinking.

• Lovable launched Lovable Cloud and AI, a platform for building full-stack applications via prompts, with built-in backend services and Gemini-powered features.

• Elon Musk announced that xAI is developing “Grokipedia,” which he describes as a major improvement over Wikipedia and a step toward better understanding the universe.

• Microsoft introduced Agent Mode in Excel and Word, along with a new Office Agent in Copilot, allowing users to create spreadsheets, documents, and presentations using natural language.

• Opera launched Neon, an AI-powered browser with agentic capabilities that can take actions on behalf of users, available through a premium waitlist.

• Meta acquired chip startup Rivos to speed up its in-house AI chip development and reduce its dependence on Nvidia.

• OpenAI reportedly generated $4.3 billion in revenue in the first half of 2025 but spent $2.5 billion on research and compute, according to The Information.

• Google unveiled Gemini-powered upgrades to its Home hardware line, including AI-enhanced Nest Cams, Doorbells, a redesigned Home app and speaker, and a new paid subscription plan.

• OpenAI’s new social app Sora quickly climbed to No. 3 on the Apple App Store, trailing only Google Gemini and ChatGPT, after launching via invite-only access.

• Hume AI launched Octave 2, a multilingual text-to-speech model supporting 11 languages with advanced features like voice conversion and phoneme editing.

• Character AI removed Disney-owned characters such as Elsa, Moana, Spider-Man, and Darth Vader after receiving a cease-and-desist from Disney.

• Pew Research found that 9% of U.S. adults are getting news from AI tools, with one-third struggling to verify accuracy and half encountering misinformation.

• Google introduced new visual search tools in AI Mode, enabling users to shop across over 50 billion product listings using both image and text input.

• Zhipu AI released GLM-4.6, an open-source model with a 200k context window that outperforms Claude Sonnet 4 and DeepSeek-V3.2 on multiple benchmarks.

Closing Thoughts

That’s it for us this week.

Make sure to pay it forward if you liked any of it !