- The Neuron
- Posts
- šŗ AI builds apps in 2 mins flat
šŗ AI builds apps in 2 mins flat
PLUS: AI content moderators sue over "psychological harm"...
Welcome, humans.
Thereās a bonkers AI video going viral on Facebook, and it needs no introduction other than the screenshot below (bet you can guess why weāre sharing it)ā¦
Thereās also a Reddit version.
According to Reddit (where the video has been upvoted 63K times), this video has over 250K likes on Facebook. So clearly, something about it is working. We think itās a āso bad itās goodā situationā¦
Now, if you want to see something AI created thatās ACTUALLY good, check this out:
Hereās the launch video (with a perfect song), the research page, and the research paper behind it.
Edify 3D converts text descriptions into detailed, production-ready 3D models in two minutes, automating what typically requires hours of manual 3D modeling work.
The end goal of all this is to create something as fluid and easy to manipulate as this example (which may remind Bladerunner 2049 fans of the memory maker).
Now imagine the above cat video made with 10x better toolsā¦ could a FB vid qualify for The Oscars? Asking for a (furry) friendā¦
Hereās what you need to know about AI today:
We explain the new AI coding assistant Lovable.
Amazon invested another $4B in Anthropic.
AI content moderators sued SAMA and ScaleAI over workplace trauma.
School won case defending punishment of student's AI use.
It was a BIG weekend for AI coding startupsā¦
Thereās a new coding startup in town, and it just MIGHT have everybody else shaking in their boots (weāll qualify that in a sec, donāt worry).
Itās called Lovable, the āworldās first AI fullstack engineer.ā It just hit #1 on Product Hunt, and the user reviews are wild. Not the polite ācongrats on your launch!āāwe're talking āfixed my friend's project deadline in 5 minutesā and ābetter than what I built in Bubble in a dayā kind of wild.
Back-up. Full stack engineer. What does that actually mean?
Think of it this way: Building a real working app needs multiple people (or one very skilled developer) to handle different parts:
Someone to design how it looks (the frontend).
Someone to make it actually work (the backend, security, payments).
Someone to deploy it all.
Lovable does all of that by itself. Tell it what you want to build in plain English, and it creates everything you need. Want users to be able to log in? One click. Need to store data? One click. Want to accept payments? You get the idea.
Early users are backing up these claims. One person even launched a startup that made Product Hunt's top 10 using just Lovable.
As for us, we made a Wordle clone in 2 minutes with one prompt. Only edit needed? More words in the dictionary. Itās like, really easy yāall.
Now, Lovable is far from the first company to come up with this. Two other major AI coding assistants, v0 from Vercel and Cursor, both dropped big updates this weekend (are they nervous?)
v0 added its own full-stack support and direct deployment features.
Cursor dropped a new AI agent that can navigate your entire codebase.
You can think of v0 as the more powerful Lovable (example: someone used v0 to make v0), while Lovable has the better user experience (and Cursorās more hands-on).
Fun fact: we were going to use todayās newsletter to make a tutorial on how to create your own app with v0 before we tested Lovable, and, well, fell in love.
There's just something different about Lovable's approach. While others are making AI coding assistants, Lovable seems to have built an AI that actually understands how to build complete, working applications.
Besides those three, thereās also Replit, Bolt.new, and the upcoming Github Spark. If you need help figuring out which one to use, thereās this guy, who literally made a tier list maker for ranking the top AI software engineersāusing Lovable, of course.
FROM OUR PARTNERS
Heads up: The deals may be sweet, but your Black Friday shopping spree is a GOLD MINE for data brokers.
Every click, purchase, and loyalty signup feeds into a massive data marketplace where your personal info gets bought and sold without your knowledge.
Think beyond spam calls and targeted adsāthis data can affect:
Your loan rates.
Insurance premiums.
Even make you a prime target for sophisticated scams.
Here's the good news: Incogni helps you take back control. Their automated service hunts down your personal data across hundreds of brokers and forces them to delete it.
Even better? You get 58% off until Black Friday with code NEURON.
Treats To Try.
*Attention listens to your calls in real-time and helps you find relevant data when theyāre overāyou can even go back and watch the relevant sections. Attention can do more, tooābook a demo and try it out for yourself here.
Cades is like Lovable, but for mobile apps; you build a mobile app by describing what you want in a chat (free to create mockup, but gotta pay for the code).
Autoflow Studio creates and runs automated software tests by recording your clicks through a website, then runs the same tests across different browsers to check if your site breaks.
Pine handles negotiates your bills, cancels your subscriptions, and fights for refunds for youāall YOU do is explain your issues.
FLUX.1 Tools let you edit and expand images, add depth and edge effects, and create image variations for the open-source FLUX image generator (available through the BFL API).
Claude now integrates with Google Docs, allowing you to easily add documents to provide more context and get more relevant assistance during your conversations and projects.
Suno V4 generates original songs and remixes with features like audio remastering, lyric writing assistance, and custom cover art generation.
*This is sponsored content. Advertise in The Neuron here.
Around the Horn.
Want to see a montage of Figure 02 robots working at BMW? Here you go (good convo in the comments too).
Amazon invested $4B into Anthropic as part of a deal that will require the maker of Claude to use Amazonās Trainum2 chip to train its models, and use the Inferentia chip to deploy them.
SAMA and ScaleAI (valued at $14B), who farm out āhuman in the loopā tasks to workers in developing countries for AI companies like Meta, are being sued by 200 digital workers in Kenya for unreasonable working conditions and psychological harm after reviewing āthe worst of the worstā content.
The parents who sued a school for discipling their kid for copying AI-generated text from Grammarly into his homework lost their case.
NEW SECTION: Technical Roundup
Believe it or not, every week thereās tons of AI news we donāt cover because, well, itās a little TOO technical. So we thought weād try out a new section to round up these finds for the nerds in the audience (us included!). Here ya go:
Deepseekās new model, R1-Lite-Preview matched (or exceeded, by some accounts) OpenAI's leading o1-preview model in certain tests (try it here)
AIMv2 from Apple helps your computer better understand what's in photos and videoāif you're building apps that need to recognize objects, analyze images, or answer questions about pictures, AIMv2 makes those tasks more accurate without requiring as much training data.
SmolLM2 gives you tiny language models that run directly on your phone or laptop, letting you chat and write without needing powerful computers or internet (135M-1.7B parameters). Thereās also SmolTalk, the dataset specifically created to train the SmolM2 models.
Recall.aiās Desktop Recording SDK lets you capture meeting recordings directly from users' computers instead of using meeting bots - so your apps can grab Zoom, Teams and Meet data without joining calls.
TĆ¼lu 3 helps you customize and train language models for specific tasks like coding or math while preserving their general abilities (read more about it here, playground to test it here).
FYI: Weāre actually debating what to call this sectionāhereās a few ideas, if you want to select your fave!
What should we call this section? |
Monday Meme
A Cat's Commentary.
Thatās all for today, for more AI treats, check out our website. See you cool cats on Twitter: @noahedelman02 |
|