OpenAI's Desktop Fusion + xAI's Aggressive Talent Raid

OpenAI builds a unified desktop app for chat, code, and browsing.

XAI deploys engineers on-site to win enterprise customers from OpenAI,

and DoorDash pays careers to generate training data for AI and robotics models.

Welcome to today's Pivot 5. I'm Robin.

And I'm Sean. Let's dig in.

OpenAI is building a desktop app that bundles chat GPT,

its codex assistant, and the chat GPT Atlas browser into a one unified interface.

No launch date yet, but the standalone chat GPT app stays available.

This is clearly about competition from Anthropic and others.

If you bundle everything into one workspace,

users are less likely to jump ship to a competitor's tool.

Yeah, it's about stickiness. Keep people in your ecosystem all day

instead of switching between apps.

I think unified AI environments are becoming the real battleground now.

Right, and the question is whether they add collaboration features

or enterprise controls next. That's where the team adoption money is.

XAI is sending engineers directly into corporate offices

to win business from OpenAI and Anthropic.

The white glove approach lended them shift for payments,

which is now ditching chat GPT for GROC after the on-site implementation work.

Wait, so XAI is competing on service?

Not just model performance? That's actually smart.

Totally.

Shift4CEO says they're phasing out OpenAI,

but keeping clawed for coding.

The company's led by Jared Isaacman,

who's now NASA Administrator and a Musk ally,

so there's definitely some relationship leverage there.

Yeah, that tracks, but if this becomes the norm,

enterprise AI sales just got way more expensive.

You're embedding engineers on site now?

That skills terribly.

True, but it raises the stakes for everyone.

OpenAI and Anthropic might have to match that level of support or lose deals.

DoorDash launched an app called Tasks

that pays couriers to record videos and complete digital tasks for AI training.

Couriers film themselves doing household chores

like loading dishwashers, folding clothes,

or they record unscripted conversations in Spanish.

So they're monetizing their gig workforce for data collection.

That's clever, actually.

Couriers get paid during downtime

and DoorDash gets a scalable data pipeline.

Yeah, they're following other platforms

that realized their contractor networks are perfect

for generating the videos, audio, and annotations

that machine learning models need.

Demand for real-world training data is huge right now.

I think we'll see a lot more gig platforms

launch similar programs.

It's a secondary revenue stream

that doesn't require new infrastructure.

Geeter Thiele's Founders Fund

is leading a funding round for Halter,

a New Zealand startup making AI-powered cow collars

at a valuation over $2 billion that doubles their previous valuation.

Wait, cow collars at $2 billion?

Yeah, the collars use sensors and algorithms

to manage livestock movement and behavior.

Founders Fund has a history of betting

on unconventional infrastructure plays

and precision livestock management is growing fast

as farms adopt automation.

I mean, labor costs on farms are brutal

and herd health monitoring is a real problem.

If this works, the addressable market is massive.

Exactly.

The valuation signals investor appetite

for AI beyond software,

especially in sectors with clear unit economics.

Terms are still being negotiated,

but this positions Halter to expand

into North America and Europe.

An AI-generated replica of Val Kilmer

will appear in a large portion of a new film

called As Deep as the Grave.

The filmmakers used publicly available

generative AI tools with archived videos,

images, and audio to create the photo-realistic likeness

with approval from Kilmer's estate.

This is different from earlier posthumous appearances

like Paul Walker in Furious 7

because it's using modern generative AI

instead of just traditional CGI.

Director Court Warhees says he hopes

it's a template for doing this ethically.

Interesting.

Studios now have accessible tools

to resurrect performances

without the expense of custom CGI,

but that raises big questions about consent frameworks

and how widely estates will license AI likenesses going forward.

Yeah, I'm not so sure this opens a door

we want fully open.

The technology is impressive,

but the implications are kind of wild.

That's today's Pivot 5.

Like and Share Pivot 5.

Thanks for listening.