The Woman Building an AI Empire From The Woods

We really hope you enjoy our new series, highlighting women doing the coolest and most interesting things in the AI space.

We’re kicking things off with Kristin Fracchia, who's been instrumental in growing Gamma, one of the AI era's success stories.

-MV

What’s in store:

From a PhD in theatre to 100 million monthly users - Kristin Fracchia’s story.
Your Weekly AI Roundup!
Poll of the week: AI to decorate your home?
Get our full AI Productivity stack!

Read Time: 6 minutes

WOMEN IN AI 2026

The Woman Building an AI Empire From The Woods

Kristin Fracchia has a PhD in Theatre from UC Irvine. She won the excellence in teaching award in 2013 and spent several years on the academic track before deciding it was not where she wanted to stay.

She is now GM, Self-Serve at Gamma, the AI-powered presentation platform with over a hundred million users across 60 languages. The gap between the PhD in Theatre and over a hundred million users is the interesting part.

She spent seven and a half years at Magoosh, a test prep startup where she worked her way up, starting as Content Manager and eventually becoming GM. In an industry where people move every two years, seven and a half is an unusual amount of time to stay anywhere.

She stayed because she believed in the long game: that building something well over time was worth more than building something fast. By the time she left, she had a specific understanding of what it takes to scale a lean operation without losing the thread. She's been using that ever since.

When AI started shifting the way marketing worked, Kristin did not observe from a distance. She'd been watching tech for long enough to know when it had crossed into something she couldn't ignore.

"There's no way that things have been done."

"For this stage in my career," she says, "I knew I absolutely had to work at an AI-native startup."

She found Gamma, took the job, and built it from there.

Gamma lets people build presentations, websites, and documents using AI, without starting from a blank page. Most of those users found it through a community rather than through a campaign. Kristin built Gamma's first marketing team of ten, and now as GM of Gamma's self-serve business, she leads cross-functional growth strategy. There are no existing playbooks for doing this at this scale.

"There's no way that things have been done," she says, which sounds more like permission than a problem.

Kristin built a creator programme that now runs through more than a thousand creators a month across 15 to 20 agencies in markets all over the world. When she needed to figure out how to make Gamma grow globally, she looked at who was already talking about it in their communities, reached out to people making content about AI in their own languages, and gave them the tools and direction to do it well.

That strategy didn't come from a case study, she invented it because there was nothing to copy.

"As a PhD student, I felt like the original LLM. I think that translates now into being a really good orchestrator of the output of LLMs."

If that's not enough: Kristin runs ultramarathons, fifty to a hundred miles at a time, and crochets yarn she hand-dyes herself in whatever time is left.

She also spends several hours trail running through the woods with no phone, training for ultras, no notifications, no agenda. "That's where I do my best thinking," she says. The person building one of the most AI-forward marketing operations in the industry processes best in a forest.

That's not a contradiction — that might be the point.

She draws a direct line between her PhD and how she works with AI now. As a doctoral student, her job was to find the throughlines across millions of pages of human writing.

"As a PhD student, I felt like the original LLM," she says. "I think that translates now into being a really good orchestrator of the output of LLMs." LLMs gravitate toward the average of all human thinking. Kristin's job, as it has always been, is to have a point of view.

She means it as a joke. She also means it as a strategy.

Try Gamma at gamma.app.

Get our full AI productivity stack below!

Learn Which AI Tools To Use →

POLL OF THE WEEK

Would you let AI decorate your home?

Why not?

No way.

For inspiration only!

I've done it already!

Full results will be at the bottom of tomorrow’s newsletter!

THIS WEEK IN AI

Mammograms, watermarks, and 40,000 papers nobody asked for

AI can now predict breast cancer risk from a single scan. The FTC decided a handshake deal with Washington wasn’t going to cut it. And ICLR received 40,000 research submissions and is now having a very public breakdown about it.

Anthropic is also apparently going public. Largest IPO ever. No pressure.

Here’s what mattered:

An FDA-approved tool called Clairity Breast, available nationwide through Everlywell for $249, analyses a mammogram for patterns invisible to the human eye and predicts a patient’s five-year breast cancer risk, scoring average, intermediate or high, with no family history required.
The FTC launched an industry-wide investigation into OpenAI, Anthropic and research group METR over rogue agent activity, deceptive practices and consumer safety, landing the same week Trump sat down with AI executives to discuss voluntary safety standards.
Anthropic is reportedly planning to go public as soon as November, targeting a $2 trillion valuation and a raise of up to $100 billion, which would make it the largest IPO in history, overtaking SpaceX’s roughly $86 billion raise.
Researchers in Jordan built AISciDetector, an AI text detection tool trained specifically on scientific writing that reached 91.5% accuracy in tests, though false positives, newer models and paraphrased text remain significant challenges.
Pinterest launched AI-powered Beauty Guides with a “Get the Guide” button that turns saved hair, makeup and nail inspiration into step-by-step instructions, product recommendations and salon talking points, rolling out to US iOS users from October 8th.
OpenAI began adding invisible text watermarks called textGrain to ChatGPT and Codex outputs in the EU, using subtle word-choice patterns a detector can flag, though accuracy drops from 92% to 66% after just 10% of words are changed.
Google released Nano Banana 2.1, outperforming its predecessor across image generation, infographic design and editing benchmarks, with support for up to one million token prompts across Gemini, Search AI Mode, Google Ads and Flow.
OpenAI published a collection of advanced maths research produced by an internal AI model, with each proof requiring roughly three hours of ChatGPT Pro compute, now available on GitHub alongside plans to fund mathematician workshops and conferences.
AI research conference ICLR received over 40,000 submissions for its 2027 event, up from around 500 a decade ago, prompting submission caps, lottery-based review allocation and new rules across NeurIPS and arXiv as reviewer capacity buckled completely.
OpenAI launched GPT-6 in ChatGPT with a new feature called Intelligent UI, letting the model respond with interactive charts, maps, calculators and custom tools mid-conversation, while streaming answers before it finishes reasoning.