Attack Capital Builds AI Companies in Thirty Days

Most founders spend a decade trying to build one company that works. Attack Capital tries to build one every month. The Brooklyn firm calls itself an AI venture studio, which is a tidy label for something stranger: a workshop where ideas are pressure-tested, staffed, and shipped as software on a calendar most teams would use to plan a product launch, not finish one.

The pitch it puts on its own front door is short. It is, in its words, "the only Y Combinator alumni founded AI venture studio designed for founders who want to build and invest with speed." Read past the phrasing and there are two businesses stacked on top of each other. One builds AI companies from nothing. The other lets people put money into the kind of startups they would otherwise never get near.

Both halves run on the same fuel: the Y Combinator network that co-founder and chief executive Kaushik Tiwari walked into as a member of the Winter 2022 batch. Tiwari, a Thiel Fellow and Columbia graduate, runs the firm with co-founder and chief technology officer Saumik Tiwari. The address is not Sand Hill Road. It is 650 Franklin Avenue, in Brooklyn.

01 / THE MODELA factory, not a fund with hobbies

A venture studio is an easy thing to claim and a hard thing to run. Plenty of firms bolt the word "studio" onto a fund and hope the association sticks. Attack Capital's version has a spine, and the spine is a schedule. The studio compresses the distance from idea to a working minimum viable product into 30 days, split across three stages it describes plainly.

The constraint is the product. Deadlines force decisions, and a month is short enough that a team cannot hide behind research forever. It also sets the studio apart from an accelerator, which coaches founders who arrive with an idea already in hand. Attack Capital more often supplies the concept, the engineers, and the early strategy, then keeps equity in what results. That is closer to co-founding than mentoring.

There is a reason the cadence keeps coming up. Speed is not just a marketing line here; it doubles as a filter. A team that cannot get a prototype in front of real users inside a few weeks is usually a team chasing an idea nobody actually wants. By putting the calendar first, the studio finds that out early and cheaply, before payroll and pride make a weak concept hard to kill. The 30-day clock is less about heroics than about failing on schedule.

The firm did not start here. It began life closer to a fund, letting people "invest in YC Demo Days" - buying diversified exposure to vetted startups coming out of recent Y Combinator cohorts, everything from AI to climate. Over time it shifted from passive investing toward full-scale incubation, describing itself as something like an AI holding company. The fund did not disappear. It became the second door into the same house.

02 / THE CUSTOMERSTwo audiences, one network

Attack Capital serves two very different people. The first is a technical founder who wants a studio to co-build an AI company with them - engineering hands, a go-to-market plan, help hiring, and a warm path into the YC ecosystem for the raise that follows. The second is an accredited investor who wants exposure to early-stage startups but has no way to source them. Venture, for most people, is a closed room. The fund is the side entrance.

That second audience is worth pausing on, because it explains the firm's origin. The hardest asset to get in early-stage investing is not capital. It is deal flow - the pipeline of good companies you are allowed to write a check into. Attack Capital took the part outsiders never see and packaged it. The interesting move was not picking winners. It was building the pipe that reaches them.

03 / THE PRODUCTSA portfolio that reads like an org chart

The studio's output is easiest to understand through what it has spun out. OpenMic is an AI phone agent that handles inbound and outbound calls across healthcare, logistics and trucking, and sales automation - the kind of work that eats a receptionist's day. PowerDialer is an AI sales dialer built to lift connect rates on outbound calling, an old problem the industry has thrown headcount at for years. HealOS points AI agents at the healthcare back office, the invisible admin layer that runs a clinic.

Alongside the companies it builds, Attack Capital has backed others in the network - among them Andromeda Surgical, which is developing AI surgical robots, and Hadrius, a compliance tool aimed at the SEC-filing grind. The sectors repeat: healthcare, legal, logistics, finance, sales infrastructure, payments. These are not glamorous corners of software. They are exactly the places where a narrow, domain-specific AI tool can replace hours of manual work.

The phone is a useful example of why the strategy fits the moment. Voice has been a stubborn interface for automation - too many accents, interruptions, and half-finished sentences for older systems to handle. Modern models changed that math, and suddenly an agent that can hold a real conversation is worth building for a trucking dispatcher, a clinic front desk, or a sales team that lives on outbound calls. OpenMic aims squarely at that opening, and PowerDialer works the same seam from the sales side.

*Figures cited by Attack Capital for its incubated startups; treat as the studio's own claims.

OpenMic — AI phone agentscalls, 24/7

PowerDialer — AI sales dialerup to 10x connects

HealOS — healthcare back officehours saved / week

Andromeda — AI surgical robotsbacked

04 / THE BUSINESSHow the money is supposed to work

The economics are two streams braided together. On the studio side, Attack Capital owns equity in the companies it builds, so a spinout that grows becomes an asset on its own books. On the fund side, it earns the usual returns of an investment vehicle by putting capital into vetted startups and sharing in what they become. The firm does not sell services to clients for a fee. It takes ownership and holds it.

That design carries a tension most studios prefer not to name. If you both build companies and invest in them, every promising idea forces a question: build it in-house, or back the founder already chasing it? Attack Capital leans into the overlap rather than pretending it away. Same network, two ways to play a bet. The discipline is deciding which door a given opportunity walks through.

It also changes what a good year looks like. A traditional fund waits, sometimes for a decade, for a portfolio to mark up. A studio can generate its own upside by shipping, but it ties up its own people and capital doing so. Attack Capital's answer is to run both timelines at once - patient money in the fund, impatient labor in the studio - so that a slow quarter on one side does not stall the whole machine. Whether a small team can sustain that balance is the real test.

05 / THE EDGEDomain plus speed against scale plus patience

Attack Capital's competitors fall into two camps. There are the AI-focused studios and holding companies - the AI Funds, Atomics, and eFounders-style builders that also manufacture startups. And there are the platforms that sell access to private deal flow, the AngelLists and SPV shops. Attack Capital is unusual for standing in both camps at once, which is either a hedge or a distraction depending on how tightly it is run.

Its stated bet is that the best vertical AI will not come from a general model getting bigger. It will come from someone who lived the workflow and built for it fast. In a market where a large model can technically do a little of everything, the advantage tilts toward the narrow tool that does one boring job completely. Domain knowledge plus a 30-day clock, the thinking goes, beats scale plus patience.

The counter-argument is straightforward, and the firm would be wise to keep it in view. General-purpose models keep absorbing capabilities that once justified a standalone product, and a narrow tool built in a month can be a narrow tool copied in an afternoon. Attack Capital's defense is distribution and domain trust rather than raw technology - being the phone system a clinic already relies on beats being the cleverest demo. In these markets, the last mile is the moat.

06 / THE PEOPLETwo founders, one surname

Kaushik Tiwari's biography does a lot of the firm's positioning for it. A Thiel Fellow and Columbia graduate who came up through YC's Winter 2022 batch, he has spent time across digital health, insurtech, fintech, and AI - the exact verticals the studio now targets. He also runs an India-based AI research lab, Swades.ai, on the side. Saumik Tiwari, the technical half, arrived by way of Dorm Room Fund and a focus on financial products for underserved populations.

The culture that follows from all this is operator-led and impatient by design. A sub-1% acceptance rate for the ideas and founders it takes on is the kind of number a selective shop likes to cite, and a 30-day cadence is not a slogan you can coast on. The team is small for the surface area it covers - a couple of dozen people spread across a studio, a fund, and a growing shelf of products.

Where Attack Capital lands over the next few years is an open question, and an honest one. Running a company factory and an investment fund from the same small team is hard to do without one starving the other. But the shape of the bet is clear enough: turn the parts of Y Combinator that outsiders admire from a distance - the speed, the network, the deal flow - into something you can either build inside or buy into. In Brooklyn, on a monthly clock.