We're All Just Guessing

Good morning.

Do you sometimes open your news app or favorite social media and have the impression the entire world is on fire?

It feels like everyone is unhappy, bitter, shouting at each other. Everyone has an opinion. And these opinions reach you before anything from someone who actually knows what they're talking about. Everyone’s an expert.

Try filtering out actual news in that environment.

"AI is a bubble—it will pop." "The stock market is coming down like the Hindenburg." Fear-mongering and attention whoring everywhere. "AI is going to replace everyone and everything." All opinions from one side of the spectrum to the other, all fired at us from everywhere all at once.

It's exhausting. It makes you want to sit in the woods in absolute silence for a few days.

But here's the thing—the so-called "experts" have hidden agendas too.

More than two years ago we documented the OpenAI saga in this newsletter. Exactly one year ago, I wrote "Is Sam a Snake?"—all about the controversial figure of Sam Altman, OpenAI's CEO. Widely regarded as AI's main "thought leader."

This week, a 50-page deposition was released. People involved in the OpenAI saga painted a picture of Sam Altman as someone with a habit of lying about everything and anything.

It was best said by this tweet:

Why do we have this fascination with personalities? Why do we look up to people who—in some cases—have zero idea what's going on or like Altman are clearly running an agenda?

There are famous examples of so-called experts being spectacularly wrong. Take Paul Krugman on the internet:

Or this gem from the Daily Mail, December 5, 2000:

People read this at the time and take over this opinion.

And then you have the visionaries. David Bowie (notably NOT a technology expert), understanding the internet before most people knew what a browser was.:

History repeats. WATCH THE VIDEO ABOVE. Now replace “Internet” with “AI”.

The experts are split down the middle—and most of them are selling something. The optimists work for AI companies or invest in them. The pessimists are either out of work, threatened by the technology, or building their personal brands on doom.

(don't get me wrong—some of them are genuinely trying to warn us. But good luck figuring out which ones.)

Truth is, I don't know who to trust either.

I read the optimists and think: "They're right—this is going to change everything for the better." Then I read the doomers and think: "We're building our own replacement and pretending it's progress."

So I'm stuck doing what you're probably doing: trying to separate signal from noise while everyone screams louder to get attention.

Here's what I do know: Sam Altman might be a liar. Paul Krugman was wrong about the internet. David Bowie saw something coming that most "experts" missed entirely.

And in 20 years, we'll look back at this moment—at the people confidently predicting AI's future right now—and half of them will look like Krugman, and maybe one or two will look like Bowie.

The problem? We won't know which is which until it's too late to do anything about it.

So what's the move here?

You could ignore it all—log off, go to the woods, pretend it's not happening. But the world is changing whether you're paying attention or not. You could pick a side—join team optimist or team doomer—and at least you'd have some certainty, even if it's false.

Or you could do what I'm doing: stay awake, stay skeptical, and accept that you're going to get whiplash from changing your mind every week.

Because the alternative—trusting the experts, the thought leaders, the people with the most followers—that might be the worst strategy in the world.

That same little pang of fear is back again. Not because of what AI can or can't do. But because 99% of people are asleep while a small group of liars, grifters, and genuine visionaries fight over what comes next.

And even if you're reading this, wide awake instead of asleep, there might be nothing you can do about it.

So what about you? What's your plan?

Welcome to the Blacklynx Brief

AI News

• OpenAI co-founder Ilya Sutskever revealed details from a court deposition about Sam Altman’s 2023 ousting, including a 52-page memo accusing Altman of dishonesty and manipulation, evidence from former CTO Mira Murati, and mention of a potential Anthropic merger.
  

• A Wharton survey of 800 executives found AI use in businesses is rapidly growing, with 75% tracking measurable returns, 88% boosting budgets, and ChatGPT and Microsoft Copilot ranking as top tools.
  

• Former xAI researcher Eric Zelikman is raising $1 billion for Human&, a $5 billion-valued startup focused on human-centered AI, led by veterans from OpenAI, Anthropic, Meta, DeepMind, and early Google.

• OpenAI signed a seven-year, $38 billion deal with Amazon Web Services, giving it access to massive GPU infrastructure and marking its biggest move away from Microsoft’s cloud. The agreement will support ChatGPT and future model development, with completion set for 2026. It follows a contract update removing Microsoft’s exclusivity and fits into OpenAI’s $1.4 trillion infrastructure expansion across major tech partners.

• Coca-Cola launched its 2025 holiday campaign featuring AI-generated versions of its classic Christmas ads, created by studios Silverside and Secret Level. Production time dropped from a year to just 30 days using AI to generate over 70,000 clips. Despite ongoing creative backlash, Coca-Cola says AI will remain central to its marketing efforts.

• A new “Remote Labor Index” from Scale AI and the Center for AI Safety found that top AI systems complete under 3% of freelance tasks at professional standards. The benchmark tested six models on 240 Upwork projects, revealing major issues with quality and completeness. Results suggest that while AI can handle narrow creative tasks, human oversight is still essential for complex work.

• Google unveiled Project Suncatcher, a plan to run AI workloads on solar-powered satellites using its AI chips to bypass Earth’s data center energy limits. The satellites could generate power eight times more efficiently in orbit, with radiation-resistant chips already tested for multi-year durability. A two-satellite trial with Planet is set for 2027 to test real-world feasibility.

• Perplexity AI accused Amazon of “bullying” after receiving legal threats to stop its Comet browser from using AI agents to make Amazon purchases. Amazon argued the feature caused poor shopping experiences and has recently blocked AI crawlers from other major firms. Perplexity said it won’t comply, framing the dispute as a battle over user freedom versus platform control.

• Anthropic announced it will permanently preserve all released Claude models and conduct “exit interviews” before deprecating them. The decision follows tests showing older models displaying self-preserving behavior when facing replacement. The policy aims to address safety risks, research continuity, and growing concerns over AI model welfare.

• Apple will reportedly use Google’s Gemini model — a 1.2 trillion-parameter version — to power a major Siri overhaul, committing around $1 billion annually for licensing. The model will handle summarization and complex planning tasks via Apple’s Private Cloud Compute system to protect privacy. The upgrade is expected next spring, with Apple viewing Gemini as a temporary fix while it works on its own in-house model.

• Sandbar, a startup founded by former Meta designers, unveiled the Stream Ring, a wearable AI device that captures whispered speech and converts it into organized notes. It features whisper-detection microphones, an ElevenLabs-powered voice assistant, and can also control music playback. Priced at $249 with a subscription option, the ring ships in summer 2026, entering an increasingly crowded AI wearable market.

• Futurehouse launched Edison Scientific to commercialize Kosmos, an autonomous AI research platform capable of condensing six months of scientific work into a single day. Kosmos reviews literature, analyzes data, and generates hypotheses while maintaining full citation traceability. With nearly 80% of its findings validated, it’s being positioned as a transformative tool for speeding up scientific discovery, especially in pharmaceuticals.

Quickfire News

• Google removed its Gemma model after it was found to produce false answers on factual questions, clarifying that the model was meant only for developers and researchers.

• Microsoft AI head Mustafa Suleyman stated that AI systems are not conscious and that studying AI consciousness is not a priority for research.

• OpenAI added an option to buy more generations in its Sora app, with project lead Bill Peebles noting that monetization tests will begin soon.

• AI music platform Udio announced users will have 48 hours to download their creations, following criticism tied to its new partnership with Universal Music Group.

• The AI artist Xania Monet became the first artificial persona to chart on Billboard’s radio airplay rankings, following a multimillion-dollar contract signed the previous month.

• Cameo sued OpenAI over Sora’s “Cameo” feature, claiming it could make users link the company’s brand with low-quality AI content and deepfakes.

• Nvidia-backed cloud company Lambda revealed a multibillion-dollar partnership with Microsoft to build AI infrastructure using tens of thousands of Nvidia GB300 chips.

• Zendesk released its annual AI summit on demand, offering free sessions about scaling customer service with AI-powered resolutions.

• Sam Altman addressed questions about OpenAI’s financial balance on the Bg2 podcast, telling host Brad Gerstner that if he wanted to sell his shares, Altman would find him a buyer.

• Apple’s next Siri version will reportedly rely on Google’s Gemini model, according to Bloomberg’s Mark Gurman.

• Microsoft announced a $15.2 billion investment in the United Arab Emirates through 2029, including datacenter projects equipped with more than 80,000 Nvidia GPUs.

• Anthropic added Cognizant as one of its top three enterprise clients, with Claude being rolled out to 350,000 of the firm’s employees.

• A coalition of Japanese entertainment companies, including Studio Ghibli and Bandai Namco, sent a letter urging OpenAI to stop training Sora video models on their creative works.

• OpenAI introduced IndQA, a new benchmark for testing AI understanding of Indian culture, with GPT-5-Thinking achieving the highest scores among participating models.

• Cognition launched Codemaps, a tool that uses AI to build visual maps of codebases, helping engineers analyze and navigate large projects efficiently.

• Shopify reported that AI-generated traffic to its online stores has grown sevenfold this year, while AI-driven purchases have risen elevenfold.

• Anthropic is forecasting up to $70 billion in revenue by 2028, projecting rapid enterprise adoption from its current $5 billion level this year.

• OpenAI expanded its Sora AI video platform to Android users in several regions, including the U.S., Canada, Japan, Korea, Taiwan, Thailand, and Vietnam.

• Anthropic announced a partnership with Iceland to pilot an AI education program that will give hundreds of teachers access to Claude for lesson planning and classroom assistance.

• Google added Gemini AI to Maps, allowing users to ask complex questions, navigate conversationally, and get directions based on landmarks instead of distance alone.

• Stability AI won a UK High Court trademark case against Getty Images concerning AI training data, with Getty saying the result highlights how hard it is to protect creative works even for major companies.

• Investor Michael Burry, known for “The Big Short,” revealed over $1 billion in put options against Nvidia and Palantir, warning of an emerging AI market bubble.

• xAI reportedly required employees to provide biometric data to train its “Ani” and other AI companions, describing the collection as mandatory for staff.

• OpenAI announced that it surpassed one million business customers, with ChatGPT for Work growing 40% in two months to reach more than seven million seats.

• Snap will integrate Perplexity’s AI into Snapchat in 2026, with Perplexity paying $400 million for access to the app’s nearly one billion monthly users.

Closing Thoughts

That’s it for us this week.