The AI Daily Brief: Artificial Intelligence News and Analysis - 5 New Startups That Explain AI's Most Important Trends

Episode Date: April 4, 2025

Five new startups mark a sharp shift in how people use and build with AI. General Agents debuts ACE, a computer autopilot that uses a mouse and keyboard to complete tasks faster than any previous system. Lindy rolls out Swarms, letting cloned agents handle hundreds of tasks simultaneously, from research to outreach. Plus Vibecode and Mocha.Brought to you by:KPMG – Go to ⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠⁠https://kpmg.com/ai⁠⁠⁠⁠⁠⁠⁠ to learn more about how KPMG can help you drive value with our AI solutions.Vanta - Simplify compliance - ⁠⁠⁠⁠⁠⁠⁠https://vanta.com/nlwThe Agent Readiness Audit from Superintelligent - Go to https://besuper.ai/ to request your company's agent readiness score.The AI Daily Brief helps you understand the most important news and discussions in AI. Subscribe to the podcast version of The AI Daily Brief wherever you listen: https://pod.link/1680633614Subscribe to the newsletter: https://aidailybrief.beehiiv.com/Join our Discord: https://bit.ly/aibreakdown

Today on the AI Daily Brief, five new startups that show off some of the most important themes and trends in AI. Before that in the headlines, an AI that actually asks you to think, the AI Daily Brief is a daily podcast and video about the most important news and discussions in AI. To join the conversation, follow the Discord link in our show notes. Welcome back to the AI Daily Brief Headlines edition, all the daily AI news you need in around five minutes. We kick off today with Anthropic who have launched a specialized version of Claude aimed at Education. The appropriately named Claude for Education is, designed to provide an AI-enhanced learning experience for college students. And basically what Anthropics going for here, and their key insight really is that while students are flocking to

AI as an educational tool, simply reading AI output isn't actually necessarily a great way to learn. Instead, this new learning mode is going to position Claude as a Socratic partner that engages in dialogue and encourages critical thinking. Benjabit writes, this approach directly addresses what many educators consider the central risk of AI in education, that tools like ChatGBT encourage shortcut thinking rather than deeper understanding. By designing an AI that deliberately withholds answers in favor of guided reasoning, Anthropics created something closer to a digital tutor than an answer engine. Now, Claude for Education is currently rolling out in pilot programs across Northeastern

University, the London School of Economics in Champlain College. The Northeastern pilot is especially ambitious, with the tool being provided across their 13 global campuses serving 50,000 students and faculty. The school had already committed to put itself at the forefront of AI enhanced education with their 2025 academic plan, which had been to be a full of the first of the front of AI-enanced plan, which makes sense as their university president, Joseph E. A.une, recently published a book on the impact of AI in education. Anthropic wrote, we believe AI has the power to fundamentally transform education for the better, but only if educators are leading the charge. We develop Claude with

powerful safeguards to ensure it remains a beneficial tool for students, faculty, and administrators. This is about as classic as it gets. Andrew Allen writes, most AI gives you fish. Claude claims to teach you how to fish. Now, of course, ultimately, a huge amount of this is going to be up to the students? Are they going to take the time to actually try to engage? Or are they just going to use the shortcut? But ultimately, the models can't solve that, just as educators can't ultimately solve that. So I think it's great to see these types of experiments and approaches that can actually help people who are willing to put in the time. Next up, a fundraising story. Text-to-app platform Replit is in talks to raise at a $3 billion valuation. Bloomberg reports the company is looking to raise around $200 million

in a round that would triple their current valuation. If the round closes, this would be a big validation that vibe coding tools are one of the hottest verticals in AI, which, by the way, little preview is something we talk about extensively in today's main episode. Repplett last raised in April of 2023, with the series B bringing in 97 million at evaluation of $1.1 billion. Any sphere, the creator of rival platform cursor is also fundraising at the moment and is reportedly looking for a $10 billion valuation. Honestly, with the current pace of growth for coding assistance, neither company should have any trouble hitting their marks. In A16Z's recent survey of AI apps, They found that coding platforms are going absolutely parabolic in terms of traffic.

RepLIT's CEO Amjad Mossad recently confirmed this wave of new users are converting into paying customers, posting an image of active paid deployments on Reddit and saying one of the smoothest exponential curves I've ever seen. Data labeling service scale AI expects to double their revenue this year, speaking of AI business. According to Bloomberg, the company achieved 870 million in revenue in 2024 and ramped that to an annualized rate of $1.5 billion by the end of the year. They expect growth to continue forecasting $2 billion of revenue for this year. year. They are also currently in talks for a tender offer that would value the company at 25 billion,

which would be an 80% bump for their valuation during their last round, which closed in May of last year. The company is styled as something of an Uber for AI training. They maintain a massive team of contractors to complete labeling tasks for companies like Microsoft and OpenAI. Finally, today, something interesting from Google, the reshuffling in their AI division continues as the leader of Notebook L.M is taking over the Gemini product. Semiport reports that Sisi Shao, who led Google's chatbot program since it was still called Bard will step down immediately. Google Labs head Josh Woodward was step into her place. Woodward oversaw the launch of Notebook L.M., which of course became Google's first breakout hit of the AI era. In a memo to staff, DeepMind CEO, Demis Hasabas said that the move would, quote,

sharpen our focus on the next evolution of the Gemini app. Now, this seems ultimately to be a continuation of Google's consolidation of AI at both the organization and product level. Last month, audio overviews were integrated as a basic feature of Gemini, helping that flagship assistant become a unified interface for Google's AI products. Over the last six months, Google has also folded the Gemini app team and the AI Studio platform into the Deep Mind Division, bringing all of AI product under the same umbrella as their research unit. Look, as someone who loves the notebook L.M. product,

I'm certainly enthusiastic about the leadership there, being moved into positions of broader authority, so let's see what they can build. For now, that is going to do it for today's AI Daily Brief Headlines edition. Next up, the main episode. Today's episode is brought to you by Superintelligent and our friends at Lindy. You've heard me talk about Lindy a couple times over the past few weeks.

They are an Agent Builder platform that can help you build agents to automate a huge variety of functions, workflows, services, from data entry to lead generation to customer service. They're basically a platform for building all of the agents that you might want in the future. We had previously had an offer where for AI Daily Brief listeners, if you email us at Agent at B-Super.a.i with the word Lindy and the title, we'll connect you to the Lindy team who can help you build a specific custom agent in a matter of days or weeks for less than $20,000 a year. It's a great way to dive in and just get your feet wet with agents. Today, however, Lindy announced something very cool, specifically agent swarms. If you've heard my doctor's strange theory

of AI agent work, this is basically that brought to life. Instead of an agent doing research on sales leads, one at a time, you could spin up a swarm of 200 Lindy agents, all doing different research and writing custom emails so that in a matter of seconds you could be sitting on 200 custom emails ready to go for each of those different leads. We've been playing around with Swarms for research, for content production, for sales, and it is very much a glimpse of the future. So again, if you are interested in Lindy, send us a note at Agent at Bsuper.a. Put Lindy in the title and we will get right back to you. Today's episode is brought to you by Vanta. Trust isn't just earned, it's demanded. Whether you're a startup founder navigating your first

audit or a seasoned security professional scaling your GRC program, proving your commitment to security has never been more critical or more complex. That's where Vanta comes in. Businesses use Vanta to establish trust by automating compliance needs across over 35 frameworks like SOC2 and ISO-2-2. Centralized security workflows, complete questionnaires up to 5X faster, and proactively manage vendor risk. Vanta can help you start or scale up your security program by connecting you with auditors and experts to conduct your audit and set up your security program quickly. Plus, with automation and AI throughout the platform, Vanta gives you time back, so you can focus on building your company. Join over 9,000 global companies like Atlassian,

Cora, and Factory, who use VANT to manage risk and prove security in real time. For a limited time, this audience gets $1,000 off Vanta at vanta.com slash NLW. That's VANTA.com slash NLW for $1,000 off. Welcome back to the AI Daily Brief. Every day at feels like a million new, cool AI startups launch. So much so that it's actually pretty rare that we dig deep and cover brand new companies on this show. And yet, over the past couple of days, I've seen a handful of companies that are on the one hand just very impressive and interesting and seemingly high potential. And additionally, really reflect and tell the story of key trends shaping AI right now. So what we're going to do today is look at these five or so new companies,

all of which have been announced literally over the last 48 hours or so. Talk about them or their new products that have been announced, but then also put them in the larger context. First up, we have General Agents. CEO Sergio Ogerer, formerly of Tesla and Google DeepMind, writes, Today I'm launching my new company General Agents and our first product, introducing Ace, the first real-time computer autopilot. Ace is not a chatbot.

Ace performs tasks for you, on your computer, using your mouse and keyboard, at superhuman speeds. Now, by the way, this is a show that's going to be much better if you watch it, because there's lots of examples going on. You should be able to watch it in the Spotify app or on YouTube. But Sherjo goes on, Ace can use all the tools on your computer. And the video shared shows it grabbing images, dropping them into Google Drive, and generally doing the types of things that a human would do.

Their preview videos also show the method of interaction. In Premiere Pro, it shows a user using the Ace Bar to, for example, slow the video clip down to 50%, split the clip once with the cut tool. and then export it. And again, you're using verbal instructions and then letting Ace actually do the work. They point out that this can be really helpful, not just for increasing your speed and efficiency, but also for tools that you're still learning how to use. There's a bunch of other things about ACE as well, but the big thing that they're hammering is that it's fast. In fact, they claim 20 times faster than competing agents, making it, as they say, more practical for everyday use.

Speed is definitely what caught people's attention with this. Replet CEO, Amjad Masad said, insane speed. Computer use models have not seen major commercial breakthrough because they're slow and expensive. This is a huge unlock. William Gus, a former research scientist at OpenAI, who's now at General Agents as well, writes ACE is the world's fastest AI computer use agent. This is fundamentally a new way to use computers. When someone asked how does it work, William responded, it takes in pixels and instructions and outputs click drag keyboard, etc. actions at 300 milliseconds round-trip latency. In other words, it's doing full screenshot image processing extremely quickly. Now, in terms of the discussion around this outside of it just being fast, some people pointed out that advances coming from startups like this mean that it's not just the big labs that are going to have all the fun in the new era.

F. Schumann and Twitter also pointed out that the expansion of human creativity to go with a theme from yesterday's show is one of the byproducts of this type of computer use tool, whereas he puts it, anyone can now operate software without even knowing how to operate it. That is a huge breakthrough and certainly wasn't on my bingo card for this year. So there are obviously a lot of themes that I think this startup represents. broadly speaking, of course, agents, the most important, most dominant category of AI advancements right now, but more specifically this idea of expanding agent capabilities through computer use. This is something that we started to see inklings of at the end of last year, and this appears to be a major jump forward on that. Obviously, if you listen to yesterday's show about the five fast-changing AI transformations that'll keep you up at night, you'll have heard me talk about the expansion

of human creativity. And as we just heard, being able to talk at software and have your agent do it with you does absolutely expand which software tools a person can use. Lastly, the fact that the speed of computer use is now exceeding humans, I think gives us a glimpse into the future where AI and agents aren't just as performant as people but are actually better at things. Now, obviously, AI is better at many things already, but this is a highly visual example of that. Another theme, which it's a little too early to explore, is whether US state-of-the-art agent capabilities can drive as much virality as recent releases from China like Manus. We might be a little constrained in this case, because general agents is only available as a research preview, not as a full product.

But still, that's something that I'll be watching. Next up, we have Lindy, and of course, while Lindy isn't new, its new Swarms feature is. I won't spend a ton of time on this because I got into it yesterday, but I think co-founder Lindy Drope actually does a great job explaining this. She writes, this AI agent turns one workflow into a multi-agent empire working in parallel, fed at one event URL, and here's what happened. It extracted every speaker's details, deep researched each person using LinkedIn, news, and PR, found and transcribed their podcast appearances, crafted emails with their exact quotes, making them impossible to ignore, identified lookalike profiles, and repeated the process on autopilot. The net result, she says,

it just sent 370 impossible to ignore personalized emails in seconds versus an entire day of manual work. We're building AI employees that run entire companies autonomously. What about someone outside the company? Michael Respusi writes, First test using Lindy loops, trying out a series of custom deep research agents using perplexity and Anthropic. Context, we're prepping for our next AI and health hack, so I want a quick update across different non-invasive imaging modalities. I texted it LFUS, FNIRS, and ultrasound, and it cooked. What would have taken one to two hours manually, 25 minutes with AI tools, took two minutes with Lindy AI. One more test from Alex Carson. They write, agent swarms let AI agents duplicate themselves

to tackle hundreds of tasks simultaneously. The new capability immediately scales what's possible with AI automation by allowing agents to divide and conquer across large data sets without complex coding or custom API connections. You provide a list of tasks, and Lindy automatically spawns duplicate agents to handle each item concurrently. Testing this feature with a YouTube to blog post workflow showed the impressive potential. I created a simple flow that triggers when new URLs are added to a spreadsheet, creates a loop that processes each URL simultaneously, transcribes each video, and generates a professionally formatted blog post, and then returns all completed posts and chat messages. And so what's the theme here? It's not just agents.

Obviously, we've already covered agents as the big dominant theme. No, what's interesting to me is this is effectively my Dr. Strange Theory of AI agent work come to life. This is not a one-to-one replacement for existing work. This is the very early innings of what you can do if you were able to hire hundreds of people instead of just one for a particular task. Those 370 personalized emails that Lindy sent as an example of how you would apply this to sales. I think we are going to see so much more of this in the next few months, and it's actually going to take the new capabilities being available before people can even come close to figuring out how to actually take full advantage.

Startup number three is called paid, and a very appropriate name it is. Manny Medina writes, I'm officially launching my new company paid. We just raised 10 million euro to solve the biggest problem no one's talking about in AI. AI agents are growing fast, but face a critical challenge, how to monetize and capture the value they create. Traditional software billing doesn't solve this problem. At paid, we're solving this problem by smart pricing that aligns with actual value

creation, real-time margin tracking to ensure profitability, clear dashboards that demonstrate ROI to customers, billing, invoicing revenue recovery taxes, and all the back office needs in a single pane of glass. With just a few lines of codes, AI builders can focus on creating amazing agents while we handle the business engine behind them. Summing it up even more clearly, their website calls this the business engine for AI agents. To me, this is reflective of the fact that agents are now officially shifting from interesting toys and novelties to actual production ready tools that are going to go in and change how we do work. Paid is an example of the critical infrastructure that is being built to enable agents to live up

to their full potential. Effectively, everything that human employees do now, every service that we have around them, is going to have some proximate version in the agent world. This is going to be an incredibly rich vein for startups in the coming years, and as you can see, is already happening with gusto right now. Next up, we're moving on from agents, at least sort of, to another huge theme of the moment, which is, of course, vibe coding. And for that, we turn to the company called, yes, vibe code.

They write, in order for billions of people to vibe code, we need a tool with the power of of cursor and the simplicity of chat chippy T. That, they argue, is the new app that they've built vibe code. Now, this comes from Riley Brown and his co-founders. Riley was and is, I think, the world's biggest AI TikToker, but over the past six months or so has really gone all in on vibe coding. He's been building things for himself, using content to come up with new theories of how creators are going to code things in the future, and this is the app that he and some friends basically built for himself to do what he wanted to do.

One of the examples he showed off was Vibecoding a mobile video game in a single prompt, and what's cool about this if you've used something like Lovable and Bolt is that this is a native mobile interface. Instead of focusing on web applications, ViveCode is starting with a focus on mobile applications. And when excited users started posting on the announcement thread that sign up wasn't working,

ViveCode basically said that they were already full up. So in terms of the themes here, obviously vibe coding in general is a big one, but the mobile integration is really fascinating as well. As Eddie Yun points out, right now that the average interaction with vibe coding is still via text, but imagine vibe coding with voice. Funny enough, when Andre Carpathy originally talked about vibe coding, he talked about how he was using Whisper to talk to his computer to do it, so in some ways this would be an actual instantiation of where the whole thing started. Vibcoats Anshnanda writes,

The best feeling in the world is when you build something and others can use it. With vibe code, we hope to enable anyone get that feeling. And of course, that is the big resonance of vibe coding with people, is this expansion of human creativity and creative capacity. And that was not the only example. One of the challenges for vibe coders is that there are a whole bunch of pieces beyond just writing the code. And some of those things remain outside of people's technical capacities. Now, of course, you can use vibe coding tools or separate LLMs to get help with those things. And all of the vibe coding tools have some version of a single button push to publish. Still, Andre Carpathy summed up a lot of people's feelings

when he tweeted about a week ago, the reality of building web apps in 2025 is that it's a bit like assembling IKEA furniture. There's no full-stack product with batteries included. You have to piece together and configure many individual services. Front end and backend, hosting, database, authentication, blob storage, email, payments, background jobs, analytics, monitoring, dev tools, secrets, etc. He continues, I'm relatively new to modern web dev and find the above a bit overwhelming. EG, I'm embarrassed to share it took me three hours the other day to create and configure a superbase with a Versel app and resolve a few errors. The second-use-ray just slightly from the getting started tutorial in the docks, you're suddenly in the wilderness. It's not even code.

It's configurations, plumbing, orchestration, workflows, best practices. A lot of glory will go to whoever figures out how to make it accessible and just work out of the box, for both humans and increasingly and especially AIs. Now, a lot of the vibe coding platforms showed up in the comments saying keep an eye out. Anton from Loveable said exactly what we're building at Loveable. One of the respondents, though, was Nicholas Chariere, who wrote, this is exactly our vision for Moka. Not glued together developer tools all made slightly easier, but a fully integrated one-stop shop. And this company, Mocha, was also just announced a couple of days ago. Nicholas writes, introducing Moca, we're incredibly excited to launch the absolute best way to build

full-stack web apps. No code, no templates, just describe what you want and watch it come to life. He continues, for too long, ideas have died because building products required too much technical knowledge. We're changing this. Describe your vision in human terms, get a complete working product in minutes. Keep iterating and adding features, then deploy in one click. All integrated. Unlike other website builders, Mocha aims to be fully integrated. You won't need to pay and connect five services to get off databases frontend hosting email, et cetera. Just one. We take a highly opinionated approach based on our experience building top web products like Pinterest and Next Door. You describe outcomes Mocha handles the rest. They even gave an example

of what you could do. Nicholas calls out Tom, a small business owner in Texas who operates an HVAC servicing company. Tom used Moca to build custom tailored web apps for each customer with personalized branding and functionality, which he said increased deal conversion customer satisfaction and velocity by 200%. And indeed, he sums up our theme here at the end, entering an era where everyone not just engineers can build personal software, the era of idea people and personal software. We couldn't be more excited. Now, I haven't had a chance to dig into Moka yet, but in terms of us looking at startups both as very cool things on their own but also as reflective of trends, it is so clear that this is where vibe coding is heading. This is absolutely the next

frontier, the thing for people who are trying to push things beyond just prototyping and into production. I think it is no mean feat to actually integrate this entire experience end to end. But I also think that this is going to be completely table stakes for these companies. It is going to be an absolute onslaught race between Bolt and Lovable and Moka, and the big beneficiary is all of us who get to build things that were never possible before. So that is the list, five startups that have come out in the past week that reflects some of AI's most important trends. Again, we had general agents who released Ace, the real-time computer autopilot, Lindy who announced their agent swarms. Basically, if you ever wanted your own army of employees, now you can have it. Paid, which is building agent

infrastructure, making it easier for agents to actually interact with financial rails, vibe code, which is taking the vibe coding world onto mobile, and MoCA, which is advancing vibe coding entirely by moving towards an end-to-end experience. AI remains the most interesting and exciting space in the world, and I'm thrilled to be able to share it with you every day. For now, that's going to do it for today's AI Daily Brief. Appreciate you listening or watching, as always, and until next time, peace.