😺SORA is here...

PLUS: Google's new quantum chip just cracked the impossible

Welcome, humans.

Next Friday (12.20) just opened up for our primary ad slot. If you want to snag it + promote your awesome brand, email us at partnerships@theneurondaily.com—we’ll give 10% OFF to the first person who asks!

Yesterday, OpenAI launched SORA, and today, it’ll launch something new. To see what it is, click this link and smash that refresh button til around 10am PST…

Here’s what you need to know about AI today:

We round-up our first impressions of SORA and its (failed?) launch.
Google's new Willow quantum chip outperformed supercomputers.
Ex-OpenAI exec launched $40M voice AI startup.
Reddit launched AI search tool.

Advertise in The Neuron here.

OpenAI “released” SORA…

They finally did it. OpenAI called our bluff, and actually released their state-of-the-art generative video model, SORA!

…JUST KIDDING!

Boooo! Can’t say the demand wasn’t there I guess…

First, impressions: Footage looks crisp, and subjects stay consistent (mostly). Here's a bear eating sushi that's weirdly good, a remarkably stable close-up of a woman's face, and two creative shorts—one vibey, one festive.

And here’s a demo of SORA’s skills, complete with rack focus + exact prompt used.

That said, our favorite might be this one—and before you say anything, we know, you feel betrayed.

The cool part is the features, tho:

Storyboard: Organize and edit sequences of videos on a timeline.
Re-cut: Finds and isolates best frames to extend scenes before/after.
Remix: Modifies elements in existing videos (replace, remove, reimagine).
Blend: Combines two videos into one seamless clip.
Loop: Creates seamless repeating videos.

You can also set style presets (“archival”, “noir”, “papercraft”) and adjust aspect ratio, resolution, duration, and # of variations.

The catch? It's locked down. Hard.

No one can make accounts right now. The EU can’t use it at all. And OpenAI’s being super selective about who can create videos with real humans (if you're lucky enough to get that access, you can't even share those to the public feed).

Here’s a riddle for ya: If SORA releases but no one can use it... is it really released?

Why so strict? SORA’s wrapped in guardrails: age detection, authenticity tracking metadata, blocks on copyrighted content and public figures, and it won't touch living artists' styles. You also need rights to anything you upload.

So how good is it really?

MKBHD got early access, and did a deep dive. His verdict? SORA’s details are impressive, but its physics and object permanence are... questionable (hence why we need world models).

His tips:

SORA shines with abstracts, title slides, and aerial shots.
Want animations? Go for stop-motion—the jerky movements actually work.
Remixing (a key feature) will take some work to make work right.
Be prepared to wait: to generate a 360p clip for 5s takes ~20 seconds; 1080p (HD quality) for 10s takes several minutes.

Other creators are split:

El.Cine says SORA’s borderline unusable (with examples).
Nick St. Pierre is impressed by the image to video, but thinks the content moderation is wack.

A completely synthetic nature documentary. Video w/ SORA, audio w/ Eleven Labs, Music w/ SUNO.

Want to try it yourself? Here’s OpenAI’s official help docs + a video tutorial. ChatGPT Plus ($20/month) gets you 50 basic 720p clips, while Pro ($200/month) unlocks 500 videos at 1080p... if you can get in!

WDYT, do you still want to try SORA?

Can't wait!

I'm OVER waiting, so no.

Don't need it, other video tools work fine.

Not interested in AI video stuff at all.

FROM OUR PARTNERS

This Smart Home Company Has $10 Million in Revenue and Just Doubled Its Monthly Growth

Smart home technology is booming, and RYSE is at the forefront:

10 granted patents.
Exclusive agreement with Amazon protecting their tech.
Hit $10M in revenue, and growing 200% month-over-month.

What do they do? RYSE transforms regular blinds into automated, voice-controlled systems, making home automation both easy and affordable.

Already distributed in 127 Best Buy locations, RYSE is preparing for international expansion and partnerships with retail giants like Home Depot and Lowe’s.

Want in? You can invest at $1.75 per share, but don’t wait—their investment round closes December 11, 11:59PM EST.

Last chance to invest before their share price change.

Treats To Try.

*Check this out: You can take your Elf on the Shelf game to new heights this holiday season with Adobe Firefly! Learn more and get inspired by the community!
Webhooks in Notion let you trigger automations directly from your documents and databases where you're already working, letting you publish content or create tickets without leaving your notes and projects.
Affiliate Plus from Lasso maximizes your Amazon affiliate earnings by automatically finding higher-commission products, managing your links, and tracking revenue (good explainer video).
Pickle creates a clone to stand in for you during video meetings, then lets you jump back in whenever you're ready (backed by NVIDIA).
NotionFlashcard transforms your Notion notes into interactive flashcards to help you remember more by testing you at the right interval.
Soundfont Generator is a weird / fun music generator where you describe a certain style of sound you want, then you type away to make music in that style.
Remy analyzes your sleep patterns, predicts your ideal schedule for activities, and coaches you through science-backed routines to optimize your daily energy (iOS only rn). There’s also Urso, who gamifies your self-care routines with a virtual bear friend.

See our top 51 AI Tools for Business here!

*This is sponsored content. Advertise in The Neuron here.

Around the Horn.

Google demo’d a new quantum chip called Willow that performed a calculation in 5 minutes that would take today's best supercomputer 10 septillion years (video).
NVIDIA is under an antitrust investigation in China over an acquisition the company made for a company called Mellanox.
OpenAI's former voice assistant researcher founded WaveForms + raised $40M from Andreessen Horowitz to make computers feel more natural to talk to by understanding the emotional tones in your voice.
Reddit Answers is Reddit’s upcoming AI search tool to help you search through Reddit conversations.
A study found that researchers using AI discovered 44% more new materials (and filed 39% more patents) than those who didn’t use AI.

FROM OUR PARTNERS

Live Webinar Next Week:

Learn to generate chunked, vectorized JSON from any document type and build accurate ETL pipelines for RAG using Unstructured + Ragas in minutes. Register to join or get the recording.

Under the Hood

Deepthought-8B is an 8B-parameter Llama-based model that documents its reasoning in JSON format, requires 16GB+ VRAM to run, and includes features like test-time compute scaling and programmable reasoning patterns (try here).
ContextCite helps you verify generated content by tracking and highlighting which specific sources the AI used to create its responses (here’s a demo, read more about it here).
MagicQuill lets you edit images precisely by letting you draw what you want to add or remove, then automatically understanding and applying your intended changes (can run locally, should work on ~8GB of VRAM).
Here’s a ~19 min tutorial for how to 2x your output coding using o1 Pro.