Fraktal

The data layer for the next generation of GTM tools

(email luka@trytimothy.com to chat)

The right GTM tool at the right time can make your company

Founders spend so much energy trying to pick one: demos, trials, dashboards, reviews, switches, etc.

By our count, Y Combinator has funded 103 AI GTM tools since 2024. Picking one is exhausting

The one made for you
Logos: the 103 YC companies (2024 to 2026) we classed as AI GTM tools
02

What if a GTM tool knew what works today?

It knew what works now on your channel

It ran on your own data

It showed what to do next, and did it for you

Don't know what works Buy a tool Numbers, no action Switch in 3 months The loop we see with founders

Could founders stop switching tools?

03

Enter Fraktal

Fraktal will put GTM know-how inside your AI agent. It will help you build tools like:

  • A tracker that shows if ChatGPT cites you
  • A LinkedIn writer that sounds like you
  • A cold email check before you send
  • An AI-writing check that flags text that sounds like AI

You will connect Fraktal to your AI agent and answer a few questions. Then your agent will build the tool you need

your AI agent
> /fraktal interview
Which channel do you need a tool for?
Which features?
Which data can you connect?
Building: AI search tracker
blocks: connectors, data providers, AI-writing check
Example
04

Why now? The build point in 2026

Bought tools

  • Built for everyone
  • Not updated with the latest strategies
  • Numbers and dashboards, but no clear action
Any GTM toolDashboard
Strategy library: last updated months ago

Built tools

  • Made for one company
  • Run on its own data, so they sound like it
  • Fast to build: Luka built a demo tool for Ceviche with Claude Code in about an hour, instead of buying Supademo
The build point: when building your own tool with AI takes less time than picking one
05

What matters most when code is cheap?

Knowing what works on each channel today, and why it works

Fraktal is not a context layer. A context layer knows one company. Fraktal will know what works across the market

Same output

Building the tool is the easy part. Knowing what to build is the hard part

06

GTM beyond human reading speed

Imagine if…

…your tool read the AI answers about your market, every day

…it remembered what worked on every channel, and why

…it knew the same day when a strategy stopped working

25%
2.5%
16 to 17 Oct27 Oct
ChatGPT answers that cite big news sites (WSJ, FT, Reuters
and others) fell from 25% to 2.5% in ten days. Same 484 buyer
questions. Same-day repeat runs moved about 2 points.
We read about 2,900 AI answers, twice. Ten startups, six AI engines. This is one change from one channel.

How long would a founder's team take to notice this change?

07

What we heard so far

Growth lead

"I won't click around in any tool if I can hook into yours and my content just gets better."

Growth lead

"For me one really important thing is that the tool has both an API and an MCP."

Growth lead

"I tried a couple of tools, I didn't like them, they don't give good overviews."

Growth lead, on writing from their own data

"I think it's mandatory… Anything else is, like, literally a disaster."

Matt · Founder, Ceviche (Luka's client)

On the demo tool Luka built him: "Perfect, this looks great, man."

Sales lead

"I think it's good that you're going in that infrastructural direction."

Sales lead

"It has to be scalable, affordable, and yet easy to integrate."

Growth lead

They asked for a demo, and for an AI-writing check.

Seven cards: one discovery call (28 Oct, growth lead and sales lead at one startup, translated from Serbian, not named). One card: a client. More founder calls are next.
08

Fraktal's value will compound in 3 ways

Brain: Fraktal will learn what works with every result it reads

Builds: the interview and the blocks will get better with every tool built

Network: advice will get better as more tools on Fraktal share their results

How well Fraktal knows what works, as more tools join
Tools built on Fraktal What Fraktal knows
Illustration, not data
09

Fraktal's objective function: maximize freshness

Fraktal will show which strategies still work today, and the results that prove it
Every strategy is a bet:
- if it still works, it stays
- if it stops working, every tool on Fraktal will know at the next daily check
Fraktal's definition
fresh·ness
/ˈfreʃ.nəs/
noun
  1. The quality of being new.
  2. Advice that still works today, checked against real results.

Experts do this by instinct. Fraktal will do it every day, and GTM experts will label why each result worked

10

What we're building: the data layer for the next generation of GTM tools

AI labs are racing to build agents that can build almost any software

But they do not know what works in GTM. GTM experts do.

If Fraktal can give every founder's AI agent what works now, on their own data, founders could own GTM tools that are faster, cheaper and stay up to date
11

Why you should not invest in this round

  1. This is an unproven idea: We do not know yet if founders want to build their own GTM tools instead of buying them. One startup has asked for a demo.
  2. We are very early: Fraktal is not live. Today we measure one channel (AI search) for ten startups. We think the network effect starts at about 1,000 tools built on Fraktal.
  3. No tested business model: We plan one flat subscription for founders. We have not tested pricing yet. Our focus is the brain and the first tools.
12

Why you will want to invest in this round

Over the next 5 to 10 years…

  1. Code will be cheap: most people will be able to build software
  2. Most companies will build their own GTM tools instead of renting them
  3. AI agents will run most GTM work (one startup we spoke with already runs 30 to 50 agents)
  4. What works on a channel will change in days, not years
  5. Text that sounds like AI will lose value. Tools that write from a company's own data will win
  6. Data on what works in GTM will be worth more than the tools built on it
Least controversial
Code gets cheap
Companies build own tools
Agents run GTM
Channels change in days
AI-sounding text loses value
What works is the asset
Most controversial
13