@kulpritt
2013 — 2014Critical Mass
2014 — 2016Amazon
2016 — 2018BMC
2018 — 2020Splunk
2020 — 2026Google
2026 — PresentApple
Tap to explore →
Dive Club asked me how we
built AI Studio from zero.
Jesse Kulp speaking with Dive Club host Ridd in a split-screen interview
Designing the Future of AI at Google ↗
I've spent most of my career
on tools for people who build.
Jesse at Google I/O
Fig 1. me, at someone else's keynote
Data pipelines, AI studios, developer platforms. The unglamorous middle layer.
At Amazon, I designed
for shared everyday tasks.
An early lesson in trust: who belongs, and who can act for whom.
Then Splunk.
This was the team.
Jesse with the Splunk team
We built Data Stream Processor — data you could read while it moved.
Building a data pipeline
used to mean flying blind.
You'd assemble the whole thing, ship it, and find out later.
So we put the real
events on screen.
—Streaming

Tap a stage to see the data there

See what the system is doing before you trust it to act.
Next, I helped create
the Gemini API experience.
gemini‑2.5‑pro temp 0.9

1from google import genai

2

3client = genai.Client()

4resp = client.models.generate_content(

5  model="gemini-2.5-pro",

6  contents="a text adventure…"

7)

Flip between what you asked and what you ship

Founding designer. The playground where developers try Gemini, then ship with the API.
You could describe a game,
then play it in the same tab.
Watching someone do that for the first time never really got old.
Sound Weaver — one of
the things people made with it.
Sound Weaver in Google AI Studio
Fig 2. shown at Google Cloud Next
And this was the team
that brought it to life.
Google AI Studio team together at the Gemini Playspace at Google Cloud Next
Fig 3. Gemini Playspace · Google Cloud Next
A lot of different disciplines, one very strange and fun booth.
Mostly, though,
design was the easy part.
  • Research changed the argument.
  • IA gave engineering rails.
  • A prototype aligned leadership.
  • One story, product and marketing.
Now I'm on Apple's
Human Interface team.
Jesse Kulp beneath the Apple Park rainbow
Thinking about what happens to work when software starts acting across it on our behalf.
The rest of this
I made because
I wanted to.
Nights and weekends. Nobody asked for any of it.
Cohort — what it might feel like
to work with agents as teammates.
Live
cohort-ui.vercel.app ↗
Brief a planner like you'd brief a colleague. Built on HAI-Harness ↗.
It started at a hackathon
and I haven't stopped poking at it.
Cohort hackathon
Fig 4. Cursor + a16z hackathon
Same team,
two ways in.
Cohort Creator on iPhone: a potter's wheel with a freshly thrown agent plush toy and a shelf of finished ones
Fig 5. Cohort Creator
Cohort Claw on iPhone: a neon arcade claw machine with five agent plush toys in the pit
Fig 6. Cohort Claw
Throw one on the wheel. Or fish one out of the claw machine.
Stratos — weather data
as a globe you can turn.
A live race hub
for Laguna Seca.
Live
laguna-seca.app ↗
Telemetry, strategy, and race state in one view. I built it for the track, not the couch.
Some things I'm
writing about.
  • Taste Was Never the Thing.Emotional connection as the human moat.
  • The Delegation Stage.Point-and-click, tap-and-scroll, and what's next.
  • The Proxy Problem.AI labs as extra-state actors.
  • Human in the Loop.When oversight becomes performative.
Happy to send these over — just ask.
That's the tour.
Thanks for stopping by.
@kulpritt
✉jesse@kulpritt.com 𝕏@kulpritt inLinkedIn
End of story · tap to start over ↺