Text is not all you need.

Modeling human intent via a multitude of rich modalities.

Signal 1 of 4
What you see on screen.

Which dashboard, which range, which number.
Signal 2 of 4
What you hear on calls.

Who asks, and where the answer goes.
Signal 3 of 4
Where you click and linger.

What you open, and what holds your attention.
Signal 4 of 4
The changes you make along the way.

The form you corrected it to last time.
Finally
Thought itself.

The signal before you say or do anything.

Screen

GrowthLast week

Signups 1,607

Audio

Monday call 0:42

Maya “Send last week’s signups in my #ops thread.”

Clicks and attention

DashboardWeekly reportNotes

Compare with the previous week.

Your previous edits

1,607, up 6%

Finally

Intent, from thought itself

What you would have to type todayAll you type with Intent

Send Maya last week’s signups.

Open the growth dashboard, set the range to last week, and read total signups, not trials.Understood from your screen
Maya asked on the Monday call, so reply in her #ops thread.Understood from your audio
Compare it with the week before, like the weekly report.Understood from your clicks and attention
One line, the change rounded to a whole percent, the way I fixed it last time.Understood from your previous edits
Finally, from thought itself

Closing the distance between thought and action.

Intent is being built to connect these signals over time, understanding both the work and the purpose behind it.

Our team connects work in multimodal learning and AI agents with experience building enterprise software.

Niccolo Fusai

Founded the generative AI team at Tesla AI. Early member of the post-training team at Physical Intelligence.
LinkedIn
Naman Goyal

Creator of Gemini Deep Research at Google DeepMind. Previously worked on multimodal learning at NVIDIA.
LinkedIn
Ani Venkateshwaran

An early engineer at SDF Labs. Helped build dbt Fusion and led product direction for Wizard, bringing research into software people use at work.
LinkedIn

Want to see Intent at work? Let’s talk.