From Meta Staff Engineer to Building the Presentation Tool That Fights PowerPoint Politics
Anmol Sood went from Meta intern to staff engineer in 5 years, scaled Instagram Threads to 100 million users in 5 days, then quit to build Alli. He spent five years watching brilliant engineers lose to better PowerPoint skills. Now he's fixing that with AI that doesn't compromise on quality or simplicity.
Every engineer dreams of scaling a product to 100 million users. Anmol Sood did that and walked away six months later.
From Meta intern to staff engineer in just 5 years, Anmol worked on Instagram Threads when it became the fastest-growing app in history, hitting 100 million users in 5 days. That's not a typo. Five. Days. But the celebration was short-lived because Threads then lost 80% of those users almost as quickly as it gained them. Performance was terrible, the app took 6 seconds to open when Twitter opened in one, and the team was scrambling to fix what wasn't supposed to scale that fast.[1]
But here's the twist. Anmol didn't leave because Threads failed. He left because he spent five years at Meta watching a pattern repeat itself. Brilliant engineers with game-changing ideas losing product reviews to people who were just better at making slides. The best ideas weren't winning. The best PowerPoint skills were. Directors and VPs had hundreds, sometimes thousands of people reporting to them. They couldn't keep all the context in their heads. So what mattered? How well you told your story in that 30-minute window you got with them.[1]
Today, Anmol runs Alli, an AI presentation tool that's trying to fix this. Not by making bad ideas look good, but by making the presentation quality baseline so high that only the actual idea matters. It's counterintuitive as hell. Most apps want you glued to them. Alli wants you out in 2 minutes with a deck that looks like you spent 5 hours on it.[1]
Key Takeaways: The Reality Behind Scaling and Building
Threads: The Fastest Growth and Fall:
- Built on Instagram's infrastructure, the only reason it survived 100 million users in 5 days
- Lost 80% of users because it took 6 seconds to open vs Twitter's 1 second
- Team morale stayed high because they knew it was a starting point, not the end product
Career Growth Has No Speed Limit:
- Derek Sivers' essay "There Is No Speed Limit" changed Anmol's perspective on career growth
- Spent Friday nights watching tech talks on YouTube instead of Netflix
- Didn't optimize for promotions, optimized for learning, promotions followed naturally
Why Presentations Matter More Than You Think:
- In big companies, your 30-minute slot with decision-makers is your only shot
- They manage hundreds of people, can't keep all context, rely heavily on how well you present
- PowerPoint skills became more valuable than engineering skills for career progression
The Three Paths to Y Combinator:
- Path 1: Be 20-25 from Stanford/MIT/Harvard (not democratic but real)
- Path 2: Have a product with real traction, even just 5 users who genuinely love it (most democratic)
- Path 3: Have specialized expertise that makes you uniquely suited for the idea
Q: You were inside the Threads launch. When did you realize this growth was insane and something was about to break?
Anmol Sood: It was pretty crazy. I just joined the team right around then so I was pretty new to that product. Very often when you launch things, they end up doing okay or worse than you expect. Very rarely in your career do you launch something and it does like 100x better than what you think it would do. Lots of people working very hard, not taking weekends off, just trying to get things together because it was so unexpected. The only reason it survived was it was built on a very strong foundation by teams working on Instagram for a long time. Instagram had that kind of scale already. If it was any other company without the backing of something like Instagram, it might have just died down and not been able to handle that kind of release. But personally working on it was super rewarding. Very rarely do you get to work on a product that people give so much immediate love to.
🔥 ChaiNet's Hot Take: Fast growth without infrastructure is a death sentence. Threads survived because Instagram already knew how to handle billions of users.
Q: Threads lost 80% of its users after that spike. What actually went wrong?
Anmol Sood: A lot of things. First, it wasn't planned to ever grow that big that quickly and as a result we had not planned for a lot of things. One big thing I was personally working on was performance. It took about 5 or 6 seconds to open Threads when you would open the app on your phone. Twitter opens in like a second. After the initial excitement settled down, people started seeing that okay, this wasn't a very polished product. While it was exciting, there was a lot of work required. Honestly, the team was aware of it. The team knew this is just a starting point, something on top of which we could build. Even when we lost most of our users, the morale did not go down because we were just excited about what else can we do, what else can we build on top of this. We knew we could not just copy Twitter, we'd have to build a better version, build better algorithms, build more innovative features. The team has done very well since then, innovated a lot and now managed to regain a lot of those users back.
🔥 ChaiNet's Hot Take: Losing 80% of users would kill most startups. At Meta, it was Tuesday. Morale matters more than metrics when you're playing the long game.
Q: You got promoted to staff engineer and left 6 months later. Was there a connection between the two?
Anmol Sood: There wasn't necessarily a link between the promotion and me leaving Meta more than that I had spent five years there. Even though Threads was exciting, everything started to feel the same. It started to feel like okay, I've done something similar to this before, it wasn't as new a challenge anymore. I knew that even if I left, the team is in great hands. Great people working on Threads, they're going to still make it a big product like they have. I was getting a lot more attracted to other problems out there which probably are not getting solved, which are not as lucky to get successful products. The world is full of problems. It's the perfect time to start solving those problems now that there was AI and there's a lot more one can do. I was fascinated to see what else is out there, what are other problems I can solve without having such a large company like Meta behind me. It's a feeling which develops over many months, sometimes years. There are definitely certain trigger points where you're like okay, now is definitely the time to leave.
🔥 ChaiNet's Hot Take: Comfort kills ambition. Anmol left when he could still leave on top, not when he had to leave because he was stuck.
Q: You went from intern to staff engineer in 5 years. For someone at a senior engineer level today, what's the path?
Anmol Sood: The first thing that had a lot of impact on me is this very short essay by Derek Sivers called "There Is No Speed Limit." Basically the point is there is no speed limit in your life in terms of how quickly you can grow. It's very different to everything else in your life so far. In school or college there is a speed limit. Most of us go through 12 years of school and then four or five years of college. As soon as you enter your career there is no speed limit. The world might try to impose those speed limits on you but you don't have to. You can grow as quickly as you really want to. A lot of those early years were very different. I remember I would still have fun, go out on Friday nights with friends, but then I'd come back and I'd want to watch tech talks on YouTube. Instead of Netflix I love binging on those tech talks, learning what else is out there, learning technologies. There was just this natural curiosity, just love for learning. The only two things are one, deep realization that you can grow quickly, there is no speed limit. And two, finding your area, what is it that is driving you. For me it was learning as quickly as possible, how can I solve bigger challenges instead of waiting 10 years. I wasn't necessarily trying to optimize for those promotions, just trying to optimize for how much I could learn.
🔥 ChaiNet's Hot Take: Friday nights on YouTube tech talks instead of Netflix is either insanity or genius. For Anmol, it was the latter.
Q: AI makes learning easier. But if you don't know what to ask, how do you even start?
Anmol Sood: You raise a very interesting point. AI is great when you know what's the question you want to ask, but very often the problem is you just don't know. It takes a certain amount of skill to even know what are the right questions to ask. That is where you have to bootstrap yourself with high quality material in that field. Instead of bootstrapping yourself with a lot of low-quality shallow content, go back to the actual resources, the best books written on marketing or sales, the best YouTuber who makes content on that. So that your mind gets bootstrapped with all those different terms, you understand all of it. A lot of it is then just about letting your mind be curious. The questions don't have to go to the questions, the questions come to you. Right now AI isn't super great at that bootstrapping part and I think it's an open problem. Hopefully there are many startups trying to solve that. I haven't found something great there but yes, in the interim, just getting the best source materials to bootstrap yourself in that field is the best and questions just come to you very naturally if you're curious.
🔥 ChaiNet's Hot Take: AI is a race car, but you still need to know how to drive. Bootstrap with the best resources first, then let AI accelerate you.
Q: You said you watched brilliant engineers lose product reviews to better PowerPoint skills at Meta. How does that happen at a company known for rigorous product culture?
Anmol Sood: It happens because people are too busy. The way it works is every few months you'll try to get a road map approved where you'll have directors, in some cases vice presidents present. These directors have hundreds of people, in some cases thousands of people working for them. They barely even know what all is going on within the team. It's just not possible. It's not that they're not great people, they're very skilled. But there is a limit to a human mind. You cannot keep all the context in. It's still important to get their approval because they are the ones who will ultimately sign off on giving your team the resources, engineers, the funding. It's very much like a startup where you need to pitch your company to VCs. Similarly in big companies you need to pitch your road map to people who are higher up. Just like a VC, they're not going to give you infinite time. You only have half an hour, in some cases an hour. That's the amount of time you have. These are not people you're working with every day. You might have seen them around, they might know your name, but you don't have that close working relationship. In all those situations it becomes so much more important to talk about your message clearly.
🔥 ChaiNet's Hot Take: Big company politics is just startup fundraising with a salary. Same 30 minutes, same high stakes, same PowerPoint anxiety.
Q: Was there an exact moment when you decided to quit and build an AI presentation tool?
Anmol Sood: When we decided to quit, I was exploring ideas with my co-founder Krishna around the same pain point but we were trying to explore it differently. We realized that to grow from here, no amount of technical skills, no amount of becoming better engineers is going to help in big companies because the problems become a lot more interpersonal, about pitching, managing conflicts, managing complexity, human complexity. The first idea we had was called Speak Well. We wanted people to become more effective speakers. We would listen to you on your meetings through a meeting bot and then analyze the way you speak. I sometimes speak very quickly or get into monologues. The idea was that AI would help me analyze that, give me feedback and help me improve. After we tried Speak Well for a few months, we realized okay, sure, we can help people improve their behaviors but that's hard. The much bigger opportunity is in the tools themselves, how do you make those tools easier to use so that people can spend time thinking about the message versus struggling with design.
🔥 ChaiNet's Hot Take: Don't fix humans, fix the tools. Behavior change is hard, better software is inevitable.
Q: You applied to Y Combinator. What was in that application that made them say yes?
Anmol Sood: At that point we were still doing the meeting assisting idea, Speak Well. But there were three things which were good about our application. One, we had a demo which showed that we could build. It was not all just talk. We had an actual demo where it could listen to your meetings and give you analysis, your speaking speed, words per minute, number of filler words. Something visible and tangible out there. Two and three, I think a lot of it was just obsessing over every word on the application. We got a lot of feedback from previous YC founders. Krishna and I spent over maybe 10, 15 hours just trying to nail every word in that application. Very often you write a lot and then time goes in just subtracting words, removing all the words which don't mean anything, staying to the point as much as possible. We thought we had a pretty solid application. The first thing we got was a message which practically said your idea is terrible but you guys seem interesting, do you have another idea? So that is what we heard back when we applied.
🔥 ChaiNet's Hot Take: YC said "your idea sucks but you don't." Sometimes the best compliment is an insult wrapped in an opportunity.
Q: You're competing with Canva, Gamma, Beautiful AI, Figma Slides. How is Alli different?
Anmol Sood: With Alli we obsess on two things, quality and simplicity, and we are not ready to compromise on any of them. What we've seen with all the competitors so far is either they make it too easy, so Gamma is pretty easy to use and we think they definitely nailed that, but very often the quality is super basic. People don't really like the slides they make on Gamma, they just do it because it's quick. With some other players like Canva or Figma slides, you can get to decent or great quality but it takes too long. You almost need to still have a lot of skills. As you're introducing more features the products are getting more and more complicated. With Alli, we were trying out these tools and we were like, is the problem getting solved? No. Does AI feel like an afterthought in every one of these products? Answer was yes. They all existed before AI, they're trying to market themselves with AI or bring in AI features but it was not core to their philosophy. We were like okay, if we really rethink about presentations from the ground up and AI is going to play a huge role, how do we make it easy without compromising on quality?
🔥 ChaiNet's Hot Take: Easy but ugly, or hard but beautiful. Alli is betting you can have both if you build AI-first, not AI-after.
Q: Why are presentations so hard when we can create hyper-realistic videos and images with AI?
Anmol Sood: Presentations are that hard right now because it's one of those rare areas where you need to combine both AI and humans together. Without AI presentations were just so hard because you need so many skills, you need to be very good at writing content, you need to be a great designer, you need to understand all those tools. What's possible with video and image editing is you just prompt the AI and you get a clip and you're mostly stitching them together. The AI is able to generate the entire frame itself or the entire clip itself. Versus in presentations, every single slide, sure the AI generates that but you also want to edit the content, you want to change various things about it. Sometimes you might have three points in the slide and you might need an image or visual which conveys this point better. Some of those things are best done by AI, some are still best done humanly. You don't even want to write a prompt. If you want to change the color of a certain text box, you just want to change that manually versus saying can you please convert this text to red. This interplay between human and AI operations is unique. Presentations have it over videos and images where you still need a lot of human input because ultimately it is your story, not the AI story. It is iterative. You don't end up with a perfect slide on your first go. It's like this dance between you and the AI to get that perfect slide.
🔥 ChaiNet's Hot Take: Video AI replaces you. Presentation AI collaborates with you. One is automation, the other is augmentation.
Q: Isn't AI making the problem worse by making bad ideas look good with polish?
Anmol Sood: In the short term potentially, in the long term no. Right now the equilibrium is that bad ideas, if people who have bad ideas but good skills to make a presentation, they already manage to get their bad ideas across. A lot of people who have good ideas don't. In the short term when only some people adopt Alli then yeah, sure, people who adopt Alli will get their ideas across more likely. But the vision is that everyone should present their idea where the presentation or design of the slide should not get in the way of your idea. In a world where everyone uses Alli or one of our good competitors, if most people use it, then basically it only gets down to the quality of the idea because the presentation is good for all of them, the design is good for all of them. Then all you're really evaluating is whether this idea is better than the other one. That's the sort of world we want where the best ideas win, the best stories win. It's not about the aesthetics or how much time did you spend aligning your text boxes on PowerPoint. Yes, in the short term with any equilibrium you might have these short-term struggles, but in the long term I think we are very aligned with what a better world looks like.
🔥 ChaiNet's Hot Take: Every technology gives early adopters unfair advantage. The goal is making that advantage universal so only merit matters.
Q: Google and big tech companies want to capture every market. Are you worried they'll get into presentations?
Anmol Sood: They will get into it, we want them to get into it, they're already getting into it. The story is not new. Being on the other side having worked for Meta, we really wanted to get into every single market. LinkedIn has jobs, we're like sure, we'll have a jobs product as well. Big companies want to try out everything, they have the resources and talent to do it. I think it gets down to a lot to focus. Are you focusing on it the most? It doesn't even need to be the entire audience because the presentation making audience is super large, founders, marketers, sales people, consultants, finance. We feel that one can always focus on a part of that audience and deliver better outcomes without worrying about big companies. The other thing is regardless of your company size, every company just has 10 smartest people. You might have many smart people but only 10 people are the smartest. If your smartest people are working on other things or they're not working on this thing, then you still have a chance. That's often what I realized happened at Meta, the best people would always just work on many different things. It was very hard to get 10 smart people to work on the same thing because smart people want to work on different things, to own bigger things. Versus building a team which is uniquely focused on just this one problem, everyone's super smart and everyone's super passionate about it. I think we have a chance. It's always a David versus Goliath battle when you're going up against big players but focus and obsession about this problem is going to help us win versus a company trying to do everything.
🔥 ChaiNet's Hot Take: Big companies have infinite resources but finite focus. Startups have finite resources but infinite obsession. Obsession wins.
Q: For someone who wants to get into Y Combinator, what should they do?
Anmol Sood: There are three paths and you should know which path is best suited for yourself. Path one is you have to be 20 to 25 years old and be from a top college like Stanford, MIT, Harvard, maybe the IITs. If you're not that profile then path one is out of your comfort zone. Path two is you need to have a product which is unique, you have some unique insights and then you have some traction. It could be just one user who really loves your product or five users who really love your product but they genuinely love your product. I like that path a lot because that is very democratic. Anyone, it doesn't really matter where you went to college or where you went to work, you can chase for this path. It should not be about your credentials, it should be about what is it that you're actually doing. Path three is when you have such specialized expertise. There are always people in YC who were previously working at NASA or SpaceX launching rockets and they're just exactly suited for that idea. Just because of the life experiences someone has had, they are the exact right fit for that idea. People should first be honest about themselves, which of the three paths are best suited for them and then try to optimize for that. For most of us it's path two where we don't have this one perfect idea we're best suited for and we haven't been to Stanford. Then basically one should just focus on the product, focus on talking to users, try to get any real traction and focus on the story of what is the clearest way to tell your story. Very often a lot of ideas have traction but when they start writing their application they just lose it. They start saying too many words, they start using too much AI. Use AI but the job of AI is to improve the quality of everything we're doing, not to replace your own thinking and produce lower quality than what you otherwise would have.
🔥 ChaiNet's Hot Take: Path 1 is privilege. Path 2 is proof. Path 3 is pedigree. Most of us have to take Path 2, which is also the most honest one.
Q: What does using AI well actually mean in practice?
Anmol Sood: The job of AI is to improve the quality of everything that we're doing. It's to improve the quality of the presentations we create, quality of code we write, the quality of writing we do, everything. Don't use it as a replacement for your own thinking and to actually produce lower quality than what you otherwise would have. That is something which I would tell everyone. I know that's high level but thinking about the right category first and optimizing for that is a good first start.
🔥 ChaiNet's Hot Take: AI is a multiplier, not a substitute. If you're lazy, it multiplies laziness. If you're thoughtful, it multiplies genius.
Q: Five years from now, what do you want people to remember about what Alli built?
Anmol Sood: Five years from now I want no one to ever dread making a presentation again. It should feel like a clarifying exercise versus a frustrating exercise. Once that happens, I feel like in the world we'll see a lot more ideas, the better ideas getting funded, the better projects getting adopted at companies, the better things happening over just the fancier things happening. That is the sort of world that we want to create. We actually want people to spend less time making presentations instead of more time. A lot of apps, even at Meta, our focus was just how can we make you spend more time. We are in the opposite direction. We want you to spend less time making presentations so that you can do other things. But that less time should give you better quality. It's not that we don't want you to go to a competitor where you might end up spending less time but the quality is going to get worse. How do we give you more time while also increasing the quality of your storytelling artifacts?
🔥 ChaiNet's Hot Take: Most apps optimize for addiction. Alli optimizes for liberation. Success is measured in minutes saved, not hours spent.
Final Thoughts: When PowerPoint Skills Matter More Than Product Skills, Something's Broken
Anmol's clarity: "We want people to spend less time making presentations so that you can do other things. But that less time should give you better quality."
The bottom line: Anmol's story is about recognizing problems everyone else has normalized. At Meta, he hit the career jackpot. Intern to staff engineer in 5 years, working on the fastest-growing app in history, getting promoted right as Threads was taking off. That's the dream trajectory everyone chases. But he saw something broken underneath the success. The best ideas weren't winning in product reviews. The best presentations were. Engineers with game-changing insights losing to people who spent more time aligning text boxes in PowerPoint.
Most people would've stayed. The money was good, the title was impressive, the work was impactful. But Anmol recognized that comfort is a trap. When everything starts feeling the same, when you're solving problems you've solved before just at different scale, that's when you stop growing. He left not because Meta was bad but because he'd extracted all the learning he could from it.
The pivot from Speak Well to Alli is textbook startup evolution. They wanted to fix human behavior through meeting analysis. Y Combinator basically said "your idea is terrible but you guys seem interesting." Instead of defending the idea, they listened. They realized changing human behavior is hard. Changing tools is inevitable. The insight wasn't wrong, just the solution. People don't need to become better speakers, they need better tools to communicate their ideas.
What's fascinating about Alli's positioning is the philosophical stance on AI making bad ideas look good. Anmol doesn't deny it. Short term, yes, early adopters of Alli will have an unfair advantage. But that's true of every technology. The printing press gave early adopters unfair advantage. The internet did too. The goal isn't preventing advantage, it's making that advantage so universal that only the actual idea matters. When everyone has access to great presentation design, you can't win on aesthetics anymore. You have to win on substance.
The three paths to Y Combinator framework is brutally honest. Path 1 is privilege, straight up. Be young, be from Stanford. Path 2 is proof, show real traction. Path 3 is pedigree, have specialized expertise. Most of us can only take Path 2, which makes it the most democratic but also the hardest. You can't fake traction. You can't fake users who genuinely love your product. That's the filter that matters.
For anyone in a comfortable job wondering if they should make the leap, Anmol's trajectory offers a blueprint. Don't leave because you hate your job. Leave because you've extracted all the growth from it. Don't leave for money, you'll make more in corporate. Leave for problems that don't get solved because they're not flashy enough for big companies to care. Leave when you realize the best version of yourself isn't possible where you are.
Q: How can people connect with you and try Alli?
Anmol Sood: You can find Alli at getalli.com. We're always iterating based on feedback, so if you're making presentations for anything that matters, VCs, product reviews, sales decks, try it out. You can also connect with me on LinkedIn or Twitter. I'm pretty active there and always happy to chat with people building interesting things or struggling with the same problems we're trying to solve.
Final words: The best career moves look insane from the outside until they work. Leaving Meta after getting promoted to staff engineer while working on the fastest-growing app in history sounds certifiably crazy. But Anmol saw what most people miss: success isn't measured by titles or scale, it's measured by how much you're still learning. When presentations matter more than products, when PowerPoint skills matter more than engineering skills, something's broken. He's fixing that not by making bad ideas look good, but by making the baseline so high that only good ideas stand out. In a world where everyone uses Alli, you can't win on slides. You can only win on substance. And that's exactly the world worth building.