Over 20,000 applications flew into Y Combinator, which ended up plucking out 282 startups for its latest batch. Now we’re getting our first look at them through Demo Day.
The first day’s demos included a healthy dose of artificial intelligence and open source, which is different from years past that were dominated by new fintech companies. That’s not entirely surprising; we were less than 10 minutes into Demo Day before we heard the phrase “Cerebral Valley.” Our favorites from Wednesday also extend to EV charger credit cards and a Snowflake for sensor data.
Now we’re focused on the standouts from day two. Is crypto back? What happened to the accountant tech stack? And are we ready to start talking about cloud marketplaces? Check out our favorites from the second day of demos below and see for yourself.
Please remember that we’re not offering investing advice or recommending anyone join or back a startup. We’re just having fun.
Stralis Aircraft
Details: Hydrogen-powered regional aircraft for commercial service.
Why it’s a fave: Aviation doesn’t have a clear path to decarbonization. Batteries are heavy, and biofuels and e-fuels still pump out pollution. So Stralis is pitching a hydrogen-electric hybrid power plant. The startup’s first prototype is a six-seater, and it plans on offering a retrofit of a Beechcraft 1900D for short-range, small-market flights. Its clean-sheet design is targeted at airlines currently operating short routes (less than 1,800 miles) with low passenger volumes, which today are covered by Embraer’s ERJ and Bombardier’s smaller CRJ models.
The company has optimistic timelines for each stage — 2026 entry into service for its Beechcraft retrofit, and 2030 for its clean-sheet design — but if it can hit those aggressive targets, it might garner serious interest from many airlines, especially in Europe.
Who picked it: Tim De Chant
Quazel
Details: Quazel combines one of edtech’s tried and tested use cases: language learning online and AI. The educational app teaches 21 different languages and offers real-time feedback through an AI tutor. It’s betting that the more users can have natural conversations when learning a language, the faster they can learn it.
Why it’s a fave: Given that I’ve been on the Duolingo beat for years now, I’m well versed in the tensions that sometimes define existing language learning apps. For example, app engagement may come at the cost of efficacy. Or 1:1 tutoring may come at the cost of, well, literal operating costs.
AI is now entering the chat, and I very much think that’s the tide that lifts all boats. The question ahead is whether Quazel’s scrappiness will help it beat out other apps as they test new tech; last month Duolingo announced a subscription tier that gives access to AI tutors, allowing users to practice real-world conversations with a bot. There is definitely some hope: Quazel landed 50,000 users in the first two days after launch.
Who picked it: Natasha Mascarenhas
PowerX
Details: Electricity and water monitors to help improve building efficiency.
Why it’s a fave: Building sensors is a pretty crowded space, but in its Demo Day pitch, PowerX hinted at a possible target market: restaurants. Outside of the push for induction hobs, I haven’t heard much about improving energy efficiency in restaurants. If the company can find a way to build a decent sales funnel for that market, it might find itself with a decent beachhead.
Who picked it: Tim De Chant
Turntable
Details: Turntable wants to make it easier for analytics teams to create and manage data pipelines.
Why it’s a fave: Turntable’s value proposition is based on two premises that ring true: that analytics teams struggle to keep up with internal demand for data pipelines, and that today’s tools are written for engineers not analysts. This reminded me right away of the rise of dbt — and indeed, Turntable’s site notes that it works with dbt Core. What role AI will play in its plan to enable “less technical analysts” to build data pipelines is still unclear to me, but five enterprise companies have already signed up for a pilot.
Who picked it: Anna Heim
Scanbase
Details: An API for diagnostic test analysis.
Why it’s a fave: If the pandemic taught us anything, it’s that when we don’t feel good, we need to know what’s wrong right away. Enter Scanbase, which is developing an API that helps medical companies analyze photographs of at-home rapid diagnostic tests and immediately return results on over 30 tests, including the flu, HIV, COVID and even pregnancy.
Why it’s bound to catch on: The CEO built a similar product that was acquired, and Scanbase has only been around for two months and is already seeing $600,000 ARR.
Who picked it: Christine Hall
FlexWash
Details: With $100,000 in annual recurring revenue, FlexWash wants to be the operating system for car washes. The tech offers support in the form of customer data tracking, SMS marketing, flex memberships and assisted upselling.
Why it’s a fave: Co-founder Karan Toor caught my attention when he said, “Do you remember the last time you went to a carwash and thought, ‘I wonder what software this place is running?’ You probably don’t because nobody ever thinks about carwash software.” The idea is delightful, yes, but is also bolstered by the fact that the co-founders are brothers who ran their family’s car wash business for four years before building a business in the category.
Who picked it: Natasha Mascarenhas
Squack
Details: Natural-language RPA (robotic process automation) tools for accountants.
Why it’s a fave: Aside from the super-cute name and logo, Squack provides software for accountants who don’t get enough love. Case in point: The founders say existing RPA tools have less than 5% penetration. It automates all those repetitive tasks so accountants can get back to all that number-crunching fun.
Who picked it: Christine Hall
Semantic Finance
Details: Financial news and analysis that uses AI to flag real-time events so that folks in the trading business can make faster and better decisions.
Why it’s a fave: The financial world moves quickly, with high-frequency traders executing something like a zillion trades in each pico-second. For mere mortals, just getting the news ahead of most folks will do. Naturally this means AI-powered synthesis of news signals fed into a financial data platform. You could see this idea coming from a mile away the moment that LLMs got better and, well, I dig it.
One of the key injustices in the world today is that I do not have a Bloomberg Terminal. There are a number of startups working to make it cheaper for people like myself to get access to similar data. But maybe this is what they should have been building. If the use of LLMs to better sort and tag news items works, it could be very cool.
Who picked it: Alex Wilhelm
Swishjam
Details: Open source tools that track page load times and overall site performance.
Why it’s a fave: Building a website is easy. Keeping it running is harder. But making sure that it runs smoothly and loads quickly can be a technical nightmare, and nothing makes people leave a site faster than slow load times. Swishjam is aiming to fix that by offering a suite of easy-to-use tools and dashboards to help catch errors.
It’s free to use for the first 5,000 page views, which could be enough to test out its claims. It’s also open source, so it can be deployed using your own site’s infrastructure. I’m always interested in the little bits and pieces that make a website run, so anything that quickly identifies problems is notable. It would be nice if Swishjam offered tools to help fix issues, too, but I’m definitely here for anything that helps me sift through pages of analytics.
Who picked it: Karyne Levy
Keeling Labs
Details: Reinforcement learning for electricity trading from grid-scale storage.
Why it’s a fave: There’s a wave of energy storage that will be hooking up to the grid in the coming decade thanks to declining battery prices and powerful incentives from the Inflation Reduction Act. Nearly all of that storage capacity will be able to respond in seconds, if not faster; far quicker than fossil fuel-fired peaker plants. That’ll give grid operators many more options. Keeling Labs uses reinforcement learning to not just predict demand, but also manage bidding and trading. If it works, it could help utilities and asset managers make the most of their storage capacity.
Who picked it: Tim De Chant
Suger
Details: It helps companies sell their products through cloud marketplaces, such as AWS Marketplace.
Why it’s a fave: I see cloud marketplaces as an increasingly relevant channel to tap into enterprise cloud budgets, so Suger caught my attention. I cringed at the mention that the team “built this exact product at Confluent,” but as long as it doesn’t come back to bite them, I suppose that’s a good proof of concept. Suger’s goal to offer “one platform to manage product listing, offer, contracts, metering and billing” across five cloud marketplaces sounds broad enough for B2B companies to consider trying it. The fact that it can support different types of pricing also seems well suited for a SaaS environment in which pricing models are getting more complex and hybrid.
Who picked it: Anna Heim
Radical
Details: Solar-powered aircraft that act as flying cell towers.
Why it’s a fave: Sound familiar? Yeah, Facebook tried this with Aquilla and failed. Is a tiny startup likely to succeed? I don’t know, but I always thought the idea was compelling but had yet to find its business model. Connectivity anywhere may be a huge new differentiator for mobile networks, and I bet satellites will be useful but expensive and congested. Why not a giant glider? It’s equally weird, but I appreciate the ambition.
Who picked it: Devin Coldewey
Vellum
Details: A software tool for developers to test and improve LLM prompts.
Why it’s a fave: If you couldn’t tell, a lot of startups are building products and services that employ large language models. But with the underlying software improving rapidly, LLM prompts could become wonky, out of date or simply not as useful over time. So, Vellum wants to help devs test prompts across models, keep a version history and generally tune ’em up so that they are at max power.
I presume that Vellum has an extensive roadmap to expand its product remit to support more of the LLM-in-use market, which is why I like it. We all know the picks-and-shovels argument when it comes to startups, that it is often more profitable to help other folks use a new tech than to build something on top of it yourself. The difference between Vellum and crypto infra providers is that there is lots more demand for the former.
The question I have is just how crowded its market space will be in a year’s time, and if this is the company to take over its nascent market.
Who picked it: Alex Wilhelm
Vitalize Care
Details: Mental health care for health care professionals aiming to reduce burnout and other consequences of systematic lack of support
Why it’s a fave: The workload increase in hospitals has been catastrophic to health care professionals’ mental and physical well-being, leading to burnout, quitting and labor shortages. They need a support system made for them, not just more of the same. I’ve seen a few companies looking into this space, and Vitalize seems like a good start with promising early metrics.
Who picked it: Devin Coldewey
Neptyne
Details: It lets users build applications from a spreadsheet.
Why it’s a fave: The idea of bridging the gap between spreadsheets and programming is appealing, and a combination of Python and AI seems a better way to actually achieve this goal than advanced Excel tricks.
According to its founders, “Neptyne behaves exactly like a spreadsheet but is secretly an alternative front end to a Jupyter Notebook that has an embedded spreadsheet engine.” Early adoption also seems to confirm it has potential: The team claims that its users have already built thousands of spreadsheets since Neptyne launched on Hacker News in February. The next thing to track will be whether its upcoming paid tiers also find product-market fit.
Who picked it: Anna Heim
Twig
Details: A vertical SaaS for the film, TV and music industries.
Why it’s a fave: I’m never surprised to learn about how backward the day-to-day processes are in some highly creative or lucrative industry, so hearing that major film and TV productions are relying on Excel and probably pen and paper for a lot of stuff makes sense. Going after an industry like this that needs and probably feels like it deserves a special solution is smart, and you know the budgets are there. Maybe it’ll mean a dozen or so less producers on any given film.
Who picked it: Devin Coldewey