Meet the latest betaworks AI startup Camp cohort
NYC's betaworks bowed the latest group of companies to go through its Camp program, nine AI startups building at the application layer
Betaworks’ Camp program is evolving alongside the emerging AI economy, and its latest cohort is another iterative shift: It builds on the last group’s focus on AI-native product experiences, but now moves into AI-powered applications that appeal specifically to enterprise, SMBs and future of work. The class of nine startups (the first of two cohorts betaworks is running this year) includes a wide-ranging field of founders and concepts, as is typical for the firm’s wild and wonderful approach. All of these participated in the official Camp demo day last week, which I unfortunately couldn’t make, but luckily my great pal and betaworks partner Jordan Crook filled me in, so here’s a rundown of the class for the first half of 2025:
Decode (https://getdecode.dev)
What is the whiteboard-to-code pipeline was pretty much instant? That’s what Decode proposes, with an IDE-integrated white boarding app that takes an alternative approach to moving up a layer of abstraction from the more established set of ‘vibecoding’ products like Cursor and Replit.
It’s an interesting approach and could find a sweet spot in terms of customer appeal among the class of front-end developers who are also design-oriented and/or visual thinkers. Plus, an approach that could eventually scale to offer a platform that best represents an approach to AI-powered coding which makes maximal use of spacial relationships as a core concept sounds like a solid differentiator.
CEO and founder Francois Laberge was previously Principal AI engineer at popular AI-based podcast editing tool Descript, and founding engineer Sriraam Raja has a masters from Harvard in the intersection of AI and education technology.
Trampoline (https://trampoline.ai)
Responding to RFPs is a massive time- and effort-sink, as anyone who has ever done one will tell you. It’s also often highly formulaic and repetitive, especially in the case of government contracts, and others in highly regulated industries. Trampoline aims to automate the RFP process, and says it can save between 30 to 80% of a company’s time as a result of doing so. This can help unlock new potential markets and customers for smaller companies that previously couldn’t compete with large firms on RFPs strictly as a result of not being able to dedicate resources to responding.
The team has ample experience in RFPs across a range of industries, and the startup has support from some of Canada’s top AI institutions, including The Vector Institute and Creative Destruction Lab.
TabTabTab (https://tabtabtab.ai)
One of the areas where I think there still remains a ton of work to do, and also a ton of opportunity to exploit, is in interaction paradigms involving AI and its use across computing flows. TabTabTab is a startup aimed squarely at that challenge, and is building a layer for everyday computing that resides above the OS and connects across all your various applications and productivity software.
I’ve definitely seen other startups and companies taking this same approach (including Highlight, a portfolio company of my previous employer OMERS Ventures) but this one has a great founding team in Gyanendra Mishra and Vasin Wongrassamee, two engineers with experience across some high-impact information and intelligence delivery organizations, including Bloomberg, Palantir and Citadel. Plus, this is an area where I think we’re in very early days, and I expect novel solutions to come from multiple vectors and surprise us in terms of how much they differ from computing paradigms in past eras.
Superposition (https://www.superposition.ai)
Perhaps it was because one of the investment areas that OMERS Ventures focuses on, but I feel like I’ve seen a huge number of AI-powered recruitment startups, and Superposition is one of those. There are some obvious reasons why recruitment is a category that seems ripe for AI-based innovation: Similar to sales, another AI hotbed, it’s incredibly time-consuming, and huge parts of it are rote and repetitive.
Superposition’s unique differentiators vs. the Mercors and Moonhubs of the world include that it’s agentic first, and also that it inherits the taste, experience and perspective of its founder and CEO Edmund Cuthbert, a longtime recruiter who hired Brex’s first engineer. Superposition is also specifically narrowing its focus to startup hiring at launch, which could offer it a unique wedge in a sea of competitors mostly aiming at larger and more established enterprise buyers.
Hopper (https://hopper.dev)
AI coding is undeniably a massive success and directionally the future of software engineering. But many of the tools currently on offer actually start to break pretty quickly when they encounter the complexity and consistency requirements of large, multi-departmental organizations where product and codebase development occurs across and between teams and projects. Hopper is an AI planning companion that helps shape the process from ideation to detailed tickets, keeping in mind both organizational context and technical audits from the perspective of the organization as a whole.
The idea and the expertise to build it come from founders Nathan Ackerman and Austin Riedel’s experience at big tech companies including Amazon, Microsoft and IBM.
Afterimage (website forthcoming)
Seeking remedies after something goes wrong with your own health is an intimidating task in the best of circumstances, and generally if its come to that, the circumstances definitely aren’t ideal. Afterimage uses AI to comb through patient data and find things that may have been missed – sometimes amounting to malpractice – and then helping with the rest of the process, including drafting complains, and matching patients with legal representation suited to their case.
It’s a very interesting approach, given the current state of AI-powered diagnostics and medial imaging analytics. Many reports are suggesting that AI analysis of scans and medical images can at times be even more effective than human diagnostics alone, which does raise the question of what that means not only for lab analytics going forward – but also for uncovering past cases where care wasn’t adequate as a result of (sometimes avoidable) human error.
JigsawML (https://jigsawml.com)
Mapping and maintaining enterprise software architecture is a wildly expansive, never-ending exercise that can feel like firefighting combined with wrangling cats. It’s another one of those areas that are absolutely essential at a certain scale, but also not anyone’s favorite job. As such, it’s also a perfect area to explore in terms of automation, and how AI can help with that beast of a task.
Founder Pracheer Gupta was a lead engineer at AWS as it built-out the precursor to modern vector databases, and then was a founding member of the Pinecone team when they set out to productize vector databases for AI applications.
NetAssist (https://netassist.live/seniors)
Using the insight that AI represents a new layer of abstraction when it comes to computing and computer interaction, NetAssist is developing voice and text agents that make using web-based technologies easier for senior citizens. Increasingly, community services, resources and connections are exclusively or optimally available through online and digital services, but these can be a major barrier to entry for an aging population.
Founder Anthony Hagouel once developed AI technology solutions for the largest commercial shrimp farms in the world (yes, really) but was driven to build NetAssist after having personal exposure to the capability gap between seniors and the online services they require within his family.
Graze Social (https://graze.social/)
Social media is in a state of relative disarray and flux compared to the past several years, with major incumbents seeing their dominant positions begin to erode, and new protocols like ATProto and ActivityPub picking up in terms of both interest and audience. But this new, more disparate and distributed world is a very different one when it comes to building consumer products, and also when it comes to monetization and building sustainable businesses on top, or else incorporating them into existing businesses as net additive contributors.
Graze offers everything you need in one place to build custom feeds on Bluesky, and tools to help grow your audience, as well as to set up a sponsored content marketplace to help with creator revenue opportunities. Founder Devin Gaffney is an experienced machine learning researcher and builder at the intersection of ML and online social activity.
And that’s the class for this first Camp of 2025! Shoutout to Jordan, John Borthwick and Jon Chin for always picking a super interesting batch of startups.