😸 Happy Thanksgiving from The Neuron!

Welcome, humans.

Happy Thanksgiving y’all! Since today is a holiday, we’re keeping things light, but since it’s been a huge couple weeks of news, we still wanted to recap everything you might have missed so you can skim the headlines while you chill on the couch waiting for dinner time. Or, if you’re cooking all day (much respect to you, the few, the proud, the holiday hosts), you can listen along with the podcast version here.

Now, for something light to kick us off: I guess folks are still not used to Sora 2 videos yet, because someone on Reddit wasn’t 100% sure if this video of a gang of turkeys teaming up to jump a thanksgiving blow-up was real or AI…

We think at this point it’s safe to assume if you see a hoard of animals show up on a ring camera, it’s totally AI.

And yes, “gang” is indeed the appropriate term for a group of turkeys, as it turns out.

Here’s what happened in AI today:

1. We break down what Ilya Sutskever “said” in the Dwarkesh interview.
2. HP cut 6,000 jobs to save $1B annually.
3. Prime Intellect released 106B-parameter open INTELLECT-3 model.
4. DeepSeekMath-V2 is out and achieved gold-level scores on IMO 2025

Advertise in The Neuron here!

What did Ilya Say? Here’s His Top 10 Takes…

DEEP DIVE: Unpacking the Ilya Dwarkesh interview

Wayyyy back in 2023, the was a very infamous moment in AI that led to the firing and re-hiring of Sam Altman. Amidst all that chaos, a legendary meme was born: “What did Ilya see?”

The origin of said meme kicked off when Marc Andreessen tweeted “Seriously though — what did Ilya see?” Then Elon Musk basically pulled a Kanye to Marc’s Taylor Swift by jumping in to boost the post and made the line faaaamous.

Now, at this point, “what Ilya saw” is pretty much solved. Depending on the exact timing, Ilya saw one or more of the following: the o1 the reasoning model, the (supposed) scaling wall, Sam being shady, and perhaps even his new idea for SSI and the billions he eventually raised to fund it. And ever since then, he basically disappeared.

This week, Ilya is back in the spotlight after over a year, with his new interview on the Dwarkesh Podcast. And this time around, everyone is debating what Ilya “said.”

Well, here are the top 10 things he had to say:

1. The “Jaggedness“ Problem: Ilya points out that current models are “jagged“—they crush PhD-level benchmarks but fail at basic tasks (like fixing a bug without breaking something else). He compares them to a student who studied 10,000 hours just to pass a test, but doesn't actually understand how to learn.
2. The “Age of Research“ is back: For the last 5 years, AI progress came from “scaling” (making models bigger). Ilya argues that pre-training data is running out, and we are returning to an era where ideas matter more than compute.
3. The “Secret“ Principle: Humans learn faster than AI. A teenager learns to drive in 10 hours; an AI needs millions of simulations. Ilya claims to know the “missing machine learning principle“ that explains this gap but refused to share it—hinting this is exactly what SSI is building.
4. “There are more companies than ideas“: Ilya's bluntest observation. Everyone's doing the same thing. The Silicon Valley mantra that “ideas are cheap, execution is everything“ breaks down when nobody's having ideas.
5. Emotions are the value function, not decoration: Ilya tells the story of a patient who lost emotional processing—he could still solve puzzles but couldn't decide anything. Took hours to pick socks. Emotions tell you when to stop thinking and act.
6. RL (reinforcement learning) now consumes more compute than pre-training: The balance has flipped. Long reasoning rollouts eat massive compute, and you get relatively little learning per rollout.
7. The goal isn't “finished AGI”—it's a superintelligent learner: Think of a brilliant 15-year-old, not an omniscient oracle. Deploy it, let it learn on the job, and merge knowledge across instances. That's the path.
8. You can't communicate AI power through essays—you have to show it: Ilya's evolving view: gradual deployment matters because seeing AI do something is fundamentally different from reading about it. The world needs to feel the AGI capability.
9. Ilya’s Timeline = 5-20 years to systems that learn as efficiently as humans and subsequently become superhuman. Wide range, but Ilya's not hedging—he thinks it's possible within that window.
10. Research taste = beauty + simplicity + brain inspiration: When asked what makes great AI research, Ilya's answer is almost aesthetic: “There's no room for ugliness.“ The best ideas feel right from multiple angles simultaneously.

For more insights, reactions, and our own personal takes on this, check out the deep dive above!

FROM OUR PARTNERS

Earn a master's in AI for under $2,500

AI skills aren’t optional anymore—they’re a requirement for staying competitive. Now you can earn a Master of Science in Artificial Intelligence, delivered by the Udacity Institute of AI and Technology and awarded by Woolf, an accredited higher education institution.

During Black Friday, you can lock in the savings to earn this fully accredited master's degree for less than $2,500. Build deep expertise in modern AI, machine learning, generative models, and production deployment—on your own schedule, with real projects that prove your skills.

This offer won’t last, and it’s the most affordable way to get graduate-level training that actually moves your career forward.

Learn More

Prompt Tip of the Day

Three prompt tips for you today across 3 domains.

Domain 1: viral videos (this is just fun, but can be applied to “visualizing” your own home rennovation).

Justine Moore shared the workflow behind those viral AI interior design time-lapses taking over X:

1. Start with an empty room image.
2. Use Nano Banana Pro to add design elements (loft, pool, statement furniture).
3. Feed to Veo 3.1 as start/end frames.
4. Prompt: “time lapse construction, fast motion as construction workers bring a tree into the living room and then construct the loft and add the couch and lights.”

Result = Hypnotic construction videos where entire rooms materialize in seconds.

Btw, Justine also shared that some folks use Etsy to sell genAI video and images, either to market their physical goods or doing genAI-as-a-service; using Etsy is a smart way to get distribution (instead of / in addition to spinning up a basic landing page that no one visits) and make a bit of money using AI.

Domain 2: Web design. Jason Zhou shared his tips for getting the most out of Gemini 3 for design work: the model requires precise instructions and is highly sensitive to prompts (old prompt engineering techniques and adding more prompts can actually hurt performance); he recommended Anthropic's 3-step process for prompt engineering—identify convergent defaults, find root causes and provide concrete alternatives, and structure guidance at the right altitude; and he demonstrated how this method can train agents to generate high-quality Excalidraw wireframe designs.

TL;DR (and how to apply this to other use-cases): Chaining specialized AI tools (one for design, one for video) beats trying to do everything in one prompt. For instance, use ChatGPT for search, Gemini for long context processes, Nano Banana for image editing, Claude Code for coding, and Veo 3.1 for video generation and you’re basically set.

Domain 3: Learning a new language with Nano Banana Pro! Here’s the prompt:

Draw a detailed [type of scene, example: pet shop] scene for me and label the English words for all objects. Labeling format: First line: [Your language, e.g. English] word Second line: Phonetic symbol (International Phonetic Alphabet IPA format) Third line: [Language you want to learn, e.g. Chinese ] translation”

This will create an image to learn the words of the items in the room as if you were learning from a pre-made cheat-sheet. Very cool concept!

For more like this, here’s a curated list of “Awesome” Nano Banana Prompts from Philipp Schmid / ZeroLu. Enjoy!

Treats to Try

*Asterisk = from our partners (only the first one!). Advertise to 600K daily readers here!

1. *Ideas move fast; typing slows them down. Wispr Flow flips the script by turning your speech into clean, final-draft writing across email, Slack, and docs. It matches your tone, handles punctuation and lists, and adapts to how you work on Mac, Windows, and iPhone. Start flowing for free today.
   1. BTW, if you want to learn how Wispr Flow works, Peter Yang just interviewed the founder Tanay Kothari so you can learn all about it; very cool!
2. Prime Intellect released INTELLECT-3, a 106-billion-parameter model trained using reinforcement learning that you can use to tackle complex math, code, and reasoning problems—open-sourcing the model, training framework, and 500+ environments so anyone can train their own models on the same stack.
3. DeepSeekMath-V2 is a theorem-proving model from DeepSeek that verifies its own mathematical reasoning step-by-step, achieving gold-level scores on IMO 2025 and CMO 2024—paper, code.
4. MimicKit lets you train robot movements in one simulator, then instantly test them in another (like switching from IsaacGym to IsaacLab) without rewriting any code.
5. SlopEvader is a browser extension that filters your Google searches to only show results published before November 30, 2022, so everything you find was created before ChatGPT existed and is guaranteed human-made. (Chrome, Firefox).
6. Searchable tracks where your brand appears when people search on ChatGPT, Claude, Perplexity, and Gemini, and shows you how you rank against competitors.
7. Exa 2.1 gives your apps the ability to search the web in under 500ms, which is 10x faster and more accurate, returning results in under 500ms because it's built from scratch, not wrapping Google like most competitors (raised $85M).
8. Stanford Agentic Reviewer reads your research paper, searches arXiv for related work, and gives you peer-review-style feedback in minutes instead of waiting 6 months for journal responses.
9. Peter Yang released a full tutorial on how to build beautiful websites in 12 minutes using Replit and Gemini 3 (IMO, you don’t need Replit, and we prefer other tools because Replit keeps you locked into its system; you can use something like SuperDesign for the design instead), but follow his advice and prompts via Build on AI Studio where you can publish or download the code, as it’s definitely worth it; his design video w/ Ryo Lu is also great too!

Around the Horn

1. Studying for finals over this holiday break and into December? OpenAI launched Study Mode in ChatGPT, an AI learning companion that uses Socratic questioning to guide you through problems rather than just giving answers, and Google launched a free one-year Gemini Pro plan for students with unlimited image uploads, custom quizzes, Deep Research, and Audio Overviews, plus new tools like Scholar Labs for research questions and new interactive images that let you tap diagrams for instant explanations.
2. Teaching this year and looking for AI tools to help with lesson planning and classroom prep? OpenAI also launched ChatGPT for Teachers with education-grade security, collaboration features, and admin controls—free for verified U.S. K-12 educators through June 2027
3. Workflow automation platform n8n announced a new version 2.0 with autosave, improved canvas, updated sidebar, plus a few more surprises and “breaking changes” requiring migration, scheduled for beta release December 8th and stable release December 15th.
4. OpenAI expanded data residency access to business customers in 10 regions including Europe, UK, US, Canada, Japan, South Korea, Singapore, India, Australia, and UAE, allowing ChatGPT Enterprise, Edu, and API customers to store data locally to meet regulatory requirements (yay!).
5. Amazon is steering developers away from external copilots and toward its own Kiro coding assistant as the company-standard way to write code.
6. An AI-industry backed super PAC is asking Trump and lawmakers to support a single federal AI rulebook that preempts tougher state-level regulations.
7. Separately, an Anti-AI super PAC has also popped up, creating a proxy war between competing super PAC networks as Silicon Valley donors pour roughly nine-figure sums into pro-industry groups while a rival camp of safety advocates looks to build its own war chest to keep the sector in check.
8. New York’s RAISE Act—one of the aforementioned state AI safety laws with tough guardrails—is under heavy fire as Trump the pro-AI super PAC lobby to kill the bill before it’s signed.
9. HP is cutting up to 6,000 jobs and ramping internal AI use in a cost-cutting plan it expects will save $1B annually by 2028.
10. Warner Music dropped its suit and struck a licensing pact that makes Suno a sanctioned AI music collaborator rather than an alleged infringer.

FROM OUR PARTNERS

While most teams are stuck on last-gen hardware, Dell Technologies shipped the future of AI development—and it's available now.

The Dell Pro Max with GB10 is the world's first desktop with NVIDIA GB10 Grace Blackwell Superchip. That means next-generation GPU architecture, 128GB of memory, and 4TB storage ready for whatever AI throws at you. Purpose-built for model training and data science, if you want to train and run your own AI models, this is for you.

Buy yourself a gift this year and start building on next-gen tech.

Thursday Trivia

One is AI, and one is real. Which is which? Vote below! (P.S, if this is cut off, click “view on web” at the top!)

A.

B.

Which is AI, and which is real?

Which is AI, and which is real? The answer is below, but place your vote to see how your guess everyone else (no cheating now!)

A Cat’s Commentary

Trivia Answer: A is AI, and B is real.