How AI is breaking the SaaS business model...

The last few weeks have been concerning

if you're a software developer and very

concerning if you're a software mega

corporation that relies on the SAS

business model. The model where your

customer never really owns your React

app, but instead rents the privilege to

click buttons forever at an 80% profit

margin. Adobe, Salesforce, Service Now,

Shopify, and many other companies

collectively saw $1 trillion wiped off

their market cap over the past few

weeks. But the big question is why? You

probably already know the answer. It's

not interest rates or accounting fraud.

It's AI. The suits just realized that if

an AI agent can swap out the work of 10

people in 10 milliseconds, they don't

need to buy 10 seats. They need zero. In

today's video, we'll look at seven

recent developments in the AI space that

support the thesis that software as a

service is headed towards a death

spiral. It is February 17th, 2026, and

you're watching the code report. Last

week, OpenAI released the Codeex app for

Mac OS, which they describe as a command

center for agents. The app itself had

over 1 million downloads in its first

week, and it provides a simple way to

handle agentic workflows in parallel

that anyone can use. And what makes that

so scary is that it means your boss no

longer needs to ask you, the developer,

to build an app, but instead can build

the app himself and then just ask you to

debug the 10,000 lines of code it

generated. But behind OpenAI's Codeex

app is the second thing you need to know

about, the Codeex 5.3 model. Not only is

it their most advanced coding model yet

that crushes the Trust Me Bro

benchmarks, but it's also 25% faster

than previous versions, and it's also

able to integrate skills now that allow

to do things like image generation,

writing, and research to handle the full

girth of responsibilities by a product

development team. That's impressive. But

OpenAI's biggest rival in the coding

space is Claude, which just released

Opus 4.6. 6. Not only is it excellent at

generating code, but now Enthropic is

trying to break into other areas like

legal analysis, financial modeling, and

a bunch of other things that they'll use

to justify expensive enterprise

subscriptions. But the big Silicon

Valley closed models are not the only

thing making progress. Alibaba just

released Quen 3 Coder Next, which is an

openweight, highly capable coding model

that gives companies the ability to host

their own serious developer brain behind

a firewall. And that kills another SAS

advantage, vendor lockin, because why

rent five different dev tools at $49 per

month when you can self-host your own

brain that rebuilds them all from

scratch for free. That's a no-brainer.

But another big release comes from ZAI

with GLM5. It's a model that targets

complex systems engineering and long

horizon agentic jobs, and its

performance approaches and sometimes

beats the best closed models in the

industry. But it's not the only open

model challenger. Miniax M2.5 has been

going viral the last couple days because

it also manages intelligence on par with

Frontier models at a fraction of the

compute price. We're getting very close

to the point of making these $200 AI

plans obsolete because models like M2.5

are making top tier reasoning feel

cheap, portable, and increasingly open

to anyone with a decent GPU instead of a

corporate expense account. It's pretty

clear that none of these AI companies

have much of a moat and the real battle

being fought right now is who can build

the best platform for autonomous code

orchestration. And Microsoft wants to be

the company that wins that battle with

their GitHub agent HQ. A GitHub was

originally designed just for code

hosting, but now it's a complete AI

agent orchestration platform. Agents can

open issues, generate branches, and

merge code when tests pass. Its project

management, QA, and DevOps automation

all rolled into one. But what about

Google? They've been relatively quiet

with Gemini releases lately. However,

Whimo, their self-driving car company,

just released the Whimo world model,

which is all about simulation and

prediction at scale. It demonstrates how

AI systems can model complex

environments, make decisions, and act

autonomously. But the thing is, when you

translate that into business software

like forecasting, logistics, risk

modeling, and operations, it starts to

make a lot of traditional SAS dashboards

that visualize these things look

obsolete. And with that, we've looked at

seven new developments in AI with one

theme. When intelligence becomes

abundant, software stops charging per

human. And when the seat dies, so does

the SAS profit margin. But even if SAS

dies, there will be new opportunities

for developers who know how to use

modern tools like Oz by Warp, the

sponsor of today's video. Warp already

brought you a modern terminal, and they

just launched Oz, a cloud platform for

coding agents. is so instead of being

stuck with one agent on your local

machine, Oz lets you run hundreds of

them in the cloud where they can make

changes across multiple repos

simultaneously. That means you could

have one agent fixing a bug from Linear,

another one updating docs from a pull

request, and a third one scanning logs

from a Graphana alert, all running at

the same time. You can quickly launch

new agents from the Oz web app or CLI,

or set them up on schedules and event

triggers. But once they're live, you can

watch them work and steer things as

needed. Try running your own agent today

at the link below and use the code

fireship for a special discount. This

has been the code report. Thanks for

watching and I will see you in the next

one.