Superagent (YC W24)

Human in the loop is so 2024. Agent in the loop is where it's at. It struck me today that the user experience we've built at Superagent (YC W24) is unlike anything I've seen before. Ismail Pelaseyed and I are using Superagent to build Superagent. But we're not doing it alone. Our agent is right there with us, contributing in ways that feel more like having a team member than just a piece of software. I am starting to see how a big company can be built without a big team.

Why are humans serving AI agents with tools? 🤔 While at the same time we have agents building advanced web applications for humans. It doesn't make sense to me. Why don't we let the agents identify gaps in their capabilities and have them build new tools to fill those needs themselves? At Superagent, we're building a workspace where AI agents actively learn and evolve. When an agent encounters a new challenge, it doesn't wait for a developer to notice and build a solution. Instead, we propose that the agent does the following: 1. Identifies the missing capability 2. Writes tests to define requirements 3. Implements and validates the new tool 4. Runs tool evals to measure performance and gather feedback This shift lets humans focus on strategic guidance and getting their problems solved, instead of coding agent tools. And before you are too quick to point out the AI safety issue, please be prepared to answer how the status quo solves that. AI safety is a problem we need to solve regardless. What capabilities would you want to see agents develop autonomously? Share examples from your work where this could make a difference 🌟

Achievement unlocked: a shitcoin was launched using the Superagent (YC W24) brand name. We have no affiliation with this 'Superagent AI' cryptocurrency project whatsoever. It appears to be an attempt to misleadingly associate our reputation with a worthless crypto scam. They have even contacted us to propose a deal, which we have obviously turned down. Be careful out there, folks!

How were the pyramids of Giza built? 🛕✨ It was either aliens, an advanced prehistoric civilization, or simply humans like us who had mastered the concept of leverage. Through ingenious systems of ramps, rollers, and counterweights, they achieved what seemed impossible. What will leverage look like in the age of AI? 🤔 As foundation models rapidly surpass all kinds of benchmarks, we are still stuck with limiting chat interfaces and scattered purpose-built apps. At Superagent (YC W24), we have a different vision for how humans should leverage AI. We're building an agent workspace where you can direct teams of AI agents that handle complex work for you, learning and evolving alongside you. Our vision is based on three key principles: - AI agents must be active learners, growing more capable with every interaction. - Autonomy beats hardcoded workflows as foundation models evolve. - Collaboration happens when agents and humans build trust through ongoing feedback and verification. We're seeing exciting early use cases in compliance, marketing, and R&D where Superagent can help tackle complex challenges that were previously out of reach. This is just the beginning. 2025 will be transformative for AI agents, and we can't wait to share more about Superagent's journey ahead. How would you want agents to help you do better, more impactful work in your field? 🥷

As we move from automation to reasoning, how do you price AI products? I asked Ismail Pelaseyed this question on our podcast. Ismail is a multi-time founder, multi-time CTO, and YC alum, currently building Superagent (YC W24) to bring AI-powered compliance to enterprises. I loved his take on AI pricing. Price AI like we price human reasoning, which means the closest value anchor we have is labor or salary. Here’s why: 1️⃣ 𝗔𝗜 𝗶𝘀 𝗺𝗼𝗿𝗲 𝘁𝗵𝗮𝗻 𝗮𝘂𝘁𝗼𝗺𝗮𝘁𝗶𝗼𝗻. The true value of AI isn’t in replacing repetitive tasks—it’s in augmenting or replicating reasoning. "Think about how humans charge for reasoning today: through salaries." 2️⃣ 𝗧𝗿𝗮𝗻𝘀𝗶𝘁𝗶𝗼𝗻𝗶𝗻𝗴 𝗳𝗿𝗼𝗺 𝘂𝘀𝗮𝗴𝗲-𝗯𝗮𝘀𝗲𝗱 𝗯𝗶𝗹𝗹𝗶𝗻𝗴. While tokens or seats might work for automation, they’re imperfect proxies for the value AI delivers. This was one of my favorite conversations. Ismail’s journey spans non-profit ed-tech, e-commerce, SaaS, and now AI. Key takeaways from our conversation: 𝟭. 𝗣𝗿𝗶𝗰𝗶𝗻𝗴 𝗔𝗜 𝘃𝗶𝗮 𝗥𝗲𝗮𝘀𝗼𝗻𝗶𝗻𝗴 — Instead of focusing on tokens or usage, think about the value AI creates in terms of human reasoning that has been replaced or augmented. 𝟮. 𝗧𝗵𝗲 𝗘𝗰𝗼𝗻𝗼𝗺𝗶𝗰𝘀 𝗼𝗳 𝗔𝗜 — Companies will have to view AI as more than a software purchase, and instead as a way to rethink organizational efficiency and structure. 𝟯. 𝗜𝘁𝗲𝗿𝗮𝘁𝗲, 𝗜𝘁𝗲𝗿𝗮𝘁𝗲, 𝗜𝘁𝗲𝗿𝗮𝘁𝗲 — Building startups, at any point, but especially now requires rapid iteration. From customer feedback to pricing strategy, competitiveness is a function of iteration speed. I could have asked Ismail questions for another two hours. Was such a fun conversation and the episode is great for founders building & pricing products the age of AI. Watch or listen to the full episode now 👇 🎥 Youtube: https://lnkd.in/eyBaWXv5 🎙️Spotify: https://lnkd.in/eBdAUhXY 🍏 Apple: https://lnkd.in/ejBnyNjD

I'm excited to announce that Superagent (YC W24) has successfully raised a seed round of $1.6 million USD! 🚀 This announcement comes a bit late, but our enthusiasm is as fresh as ever. Superagent aims to revolutionize web research by replacing manual efforts with AI. Our mission is to enable businesses to leverage the vast amount of information scattered across the web (and beyond) 🔍💡 We are proud and humbled to be backed by some of the best founders and operators on the planet, including: - Dharmesh Shah, founder of HubSpot - Frederic Kerrest, founder of Okta (shoutout to 515 Ventures) - Amjad Masad, founder of Replit - Rebel Fund (partners include founders of Instacart, Reddit, Inc., Optimizely) - Y Combinator (shoutout to Nicolas Dessaigne, founder of Algolia) We invite you to try the latest version of our product for free. Link in the comments.

Demo of the web research agent Superagent (YC W24) trying to guess who Clay customers are. It even spits out the apollo link so you can check who these people are. see link in description

This is my Y Combinator application. I know there is a giant mouse in the middle, but I can't reshoot it again, I have already shot this video an unhealthy amount of time haha. Bluewind is the platform to find the open-source AI worfklows to grow your business. You come into bluewind, you pick the automation you need, you set it up, you leave to go build your product. Here is an example, which is what I sent in my demo: - you put the link of your newsletter. - bluewind automagically finds your ideal audience and sends them a cold email inviting them to join the newsletter - YAY! more people benefit from your newsletter! We're working with the Bluewind stack, made of the best GTM vendors. Here are the main ones: - HubSpot -> CRM: I have done a lot of research. I know this is the mainstream guy. But there is a reason Hubspot is winning. - Smartlead -> email - HeyReach -> linkedin - LeadMagic -> data - Superagent (YC W24) for web research - All the automations are built with Windmill, the best low-code platform to ship production-ready software. And we're going to look for more vendors for twitter, for SEO, Ads, etc, etc... Again, we don't need more SaaS or more tools, we need more tool users.

Let's be honest, web research can often be tedious! Especially when you need to find specific key information across a list of websites. It's like looking for a needle in a haystack 🪡 And this challenge isn't unique to Sales; it extends to research-intensive industries like Finance and Consulting as well. But guess what...Superagent (YC W24) is here to save us thousands of hours! Its AI agents are designed for web research, making this manual task quick and easy. Companies like Algolia are already using its API for advanced workflows that require web/document research. For example, imagine you have a list of 200 LinkedIn profiles and you need to find the school for each one. You can accomplish this in 1 minute with Superagent on Airtable. Kudos to Alan Zabihi and Ismail Pelaseyed! (and Nicolas Dessaigne for the investment) -- Let's connect if you want to learn more about outbound, sales ops and tools! 🤝 Drop a like for the support 👍

This is my Y Combinator application. I know there is a giant mouse in the middle, but I can't reshoot it again, I have already shot this video an unhealthy amount of time haha. Bluewind is the platform to find the open-source AI worfklows to grow your business. You come into bluewind, you pick the automation you need, you set it up, you leave to go build your product. Here is an example, which is what I sent in my demo: - you put the link of your newsletter. - bluewind automagically finds your ideal audience and sends them a cold email inviting them to join the newsletter - YAY! more people benefit from your newsletter! We're working with the Bluewind stack, made of the best GTM vendors. Here are the main ones: - HubSpot -> CRM: I have done a lot of research. I know this is the mainstream guy. But there is a reason Hubspot is winning. - Smartlead -> email - HeyReach -> linkedin - LeadMagic -> data - Superagent (YC W24) for web research - All the automations are built with Windmill, the best low-code platform to ship production-ready software. And we're going to look for more vendors for twitter, for SEO, Ads, etc, etc... Again, we don't need more SaaS or more tools, we need more tool users.