AI Just Entered Its Sci-Fi Era

Welcome back apprentices! 👋

Hey {{first_name|friend}},

AI used to have one job: 

answer our questions without making things weird.

Apparently, that wasn’t ambitious enough. 

Now it’s wandering into labs, tackling problems humans have wrestled with for decades, and getting very good at eye contact.

So… what exactly happened while we weren’t looking?

_{Read Time: 4 minutes}

🕯️AI’s Getting a Spiritual Adviser. Anthropic co-founder Chris Olah spent the past year consulting religious scholars on a question Silicon Valley apparently wasn’t complicated enough already: Could Claude be conscious? The discussions are helping shape Claude’s morals, even as Pope Leo XIV argues AI lacks humanity’s “spark” and Sam Altman warns that giving models religious authority could become a safety problem. As AI labs start building increasingly similar products, Anthropic’s willingness to seriously entertain machine consciousness is becoming a surprisingly big philosophical dividing line.

🚨 OpenAI’s Safety Lead Walks. Safety lead David Robinson quit after 3.5 years, calling the company’s culture “broken” and warning that AI safety can’t rely on trial and error anymore. The architect of OpenAI’s current safety framework says teams were moving so fast they rarely had time to make bigger safety improvements — comparing the safeguards AI labs need to those used in airports and nuclear plants. His departure adds to a growing list of insiders raising concerns about whether AI development is moving faster than its guardrails.

🧬 AI Proteins Get Watermarked. Google DeepMind unveiled SynthID Bio, an experiment that hides detectable watermarks inside AI-designed proteins — basically a tiny “made by AI” signature for biology. Tests found the marked proteins still worked normally and could be identified after being physically produced, potentially helping DNA suppliers flag unfamiliar designs without slowing legitimate research. It’s not tamper-proof yet, but as AI makes designing biology easier, knowing where a design came from could become an important safety net.

For its first CTV campaign, Jennifer Aniston’s DTC haircare brand LolaVie had a few non-negotiables. The campaign had to be simple. It had to demonstrate measurable impact. And it had to be full-funnel.

LolaVie used Roku Ads Manager to test and optimize creatives — reaching millions of potential customers at all stages of their purchase journeys. Roku Ads Manager helped the brand convey LolaVie’s playful voice while helping drive omnichannel sales across both ecommerce and retail touchpoints.

The campaign included an Action Ad overlay that let viewers shop directly from their TVs by clicking OK on their Roku remote. This guided them to the website to buy LolaVie products.

Discover how Roku Ads Manager helped LolaVie drive big sales and customer growth with self-serve TV ads.

The DTC beauty category is crowded. To break through, Jennifer Aniston’s brand LolaVie, worked with Roku Ads Manager to easily set up, test, and optimize CTV ad creatives. The campaign helped drive a big lift in sales and customer growth, helping LolaVie break through in the crowded beauty category.

Learn more.

AI apparently got bored of writing emails and generating pictures. 

Now it’s tackling unsolved math problems, learning to simulate human cells, and showing up to video calls looking suspiciously human.

OpenAI has released hundreds of AI-generated mathematical papers, while Mark Zuckerberg and Priscilla Chan’s Biohub is backing a $1.8 billion effort to build virtual models of living cells. Meanwhile, Tavus has unveiled an AI video model so convincing that nearly half its test participants believed they were speaking to a real person.

These are three very different breakthroughs, but they point toward the same shift: AI is moving beyond answering questions to exploring science, predicting reality, and interacting with us in increasingly human ways.

Biohub is turning biology into one of AI’s biggest new experiments. 

Its expanded $1.8B Virtual Biology Initiative aims to build a “universal virtual cell” — AI capable of predicting how cells will react to a drug, disease, or genetic change before scientists run the real experiment.

The effort combines Biohub with the U.S. government, Google DeepMind, Meta, and Isomorphic Labs. But biology makes ChatGPT look easy: Biohub estimates truly accurate models could require data from trillions of cells, compared with hundreds of millions in today’s largest datasets.

If virtual cells become reliable, researchers could test ideas digitally first, narrow down the winners, and spend precious lab time on the experiments most likely to work. That could make early drug discovery dramatically faster — though virtual predictions still need real-world proof.

OpenAI dropped 722 AI-generated math papers spanning 372 groups of results from an unreleased internal model — including claimed progress on a weaker version of the Riemann Hypothesis, one of mathematics’ most famous unsolved problems.

The wild part is the speed. OpenAI says nearly all the new results came from a single prompt and roughly three hours of compute on average. And 162 papers include results written in Lean, software that lets computers check proofs step by step — though mathematicians still need to judge whether the broader claims are genuinely new and significant.

Math research could start scaling with computing power in a way it never has before. And AI wouldn’t replace mathematicians so much as give them a research partner capable of exploring an absurd number of ideas while they focus on deciding which ones actually matter.

AI avatars have officially graduated from creepy lip-sync demos. Tavus’ new Griffin model can see and hear you during a live video call, react while you’re speaking, nod naturally, handle interruptions, and even reference things visible on your screen.

And apparently, it’s getting convincing. In a small company-run study, 48% of participants believed Griffin-Lite was a real person, compared with just 2.4% for previous models.

That could mean incredibly natural AI tutors, customer-service agents, or digital assistants. It could also mean a scammer no longer needs to send you a suspicious text — they could potentially look you in the eye.

AI isn’t just learning to talk like a human anymore. It’s learning the tiny visual and social cues that make an interaction feel human — which makes knowing who, or what, is on the other side of the screen much more important.

You don't need to be a mathematician, biologist, or software engineer to feel the impact of these developments.

Currently, the most exciting possibilities involve faster medical discoveries, more capable digital assistants, and AI tools that help experts solve problems previously considered too difficult or time-consuming.

For businesses, these advances point toward new opportunities in research, customer service, education, and product development.

But they also introduce a new challenge: knowing when to trust AI-generated results, predictions, and interactions.

A convincing video conversation doesn't guarantee you're speaking to a person. A mathematical paper generated by AI isn't automatically correct. And a virtual experiment isn't a substitute for testing in the real world.

The common thread is that AI's capabilities are expanding faster than our traditional methods of evaluating them.

Learning to distinguish impressive demonstrations from reliable, independently validated technology may become one of the most valuable skills of the AI era.

Hey, you didn’t get all this info for nothing — share it! If you know someone who’s diving into AI, help them stay in the loop with this week’s updates.

Sharing is a win-win! Send this to a friend who’s all about tech, and you’ll win a little surprise 👀

Aligned shows you what buyers are doing between meetings inside your most important deals then surfaces the risks, openings, and next steps. No more guessing where a deal stands or finding out a champion went quiet too late. Just real visibility, so you can act before it's too late.

Start Using Aligned Free

A playground for your imagination (and low-key prompt skills).

Each send, we give you a customizable DALL·E prompt inspired by a real-world use case — something that could help you in your business or job if you wanted to use it that way. But it’s also just a fun creative experiment.

You tweak it, run it, and send us your favorite. We pick one winner to feature in the next issue.

Bonus: you’re secretly getting better at prompt design. 🤫

Last week, we challenged you to test GPT-4o’s visual generation skills with this prompt.

Here’s the WINNER:

Congrats to Jose for this creation!🥳

Want to be featured next? Keep those generations coming!

Deep inside a gigantic greenhouse powered by advanced robotics, [your object] has become the central component of an extraordinary living ecosystem. Mechanical arms tend to towering tropical plants, tiny irrigation drones weave between enormous leaves, and transparent pipes carry glowing turquoise water through the structure. The object remains instantly recognizable, but its surfaces are integrated with botanical details, precision-engineered components, and vivid bioluminescent accents. Sunlight pours through the glass ceiling, illuminating emerald foliage, bright coral flowers, and polished metallic surfaces. Ultra-detailed photorealism, cinematic wide-angle photography, vibrant colors, crisp textures, dramatic natural lighting.

If AI outputs feel inconsistent, it’s usually not the model, it’s missing structure.
We documented the exact 6- Part System we use to get reliable results across ChatGPT, Claude, and Gemini.

It’s a short guide you can finish in under an hour, with plug-and-play prompts + exercises so you actually build the skill and fix the frustrating AI inconsistencies.

Subscriber Price: $10 (normally $19).

_FEEDBACK