Why Modern Creator?

Christian Peverelli · YouTube

OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

A plain-English walk through OpenAI's DevDay: DOTS, GPT-6.1 Sol, Ultrafast, a full ChatGPT office suite, and which kinds of startups are now most exposed.

Posted

September 30th

today

Duration

08:51

Format

Talking Head

educational

Views

1.4K

51 likes

Big Idea

The argument in one line.

OpenAI's DevDay packed more than 20 releases into under an hour, and the ones that matter most, an always-on agent, a cheaper flagship model, and a built-in office suite, threaten any startup that's just a thin wrapper around ChatGPT.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

You build or run a SaaS or app on top of ChatGPT or OpenAI's API and need to know what changed overnight.
You want a fast, jargon-free rundown of a major AI platform announcement without watching the full keynote.
You're deciding whether to build a note-taking app, workspace tool, or AI wrapper and want to see what OpenAI just shipped natively.

SKIP IF…

You already watched the full DevDay keynote or read OpenAI's own release notes.
You want deep technical benchmarks on GPT-6.1 Sol rather than a plain-English overview.

TL;DR

The full version, fast.

OpenAI's DevDay ran about 50 minutes and packed in more than 20 releases, led by DOTS, an always-on agent that gets its own cloud computer and browser, connects to 4,000+ apps, and works while you're away. The other headline release is GPT-6.1 Sol, a model OpenAI claims beats Claude 5.5 Opus at roughly a third of the cost of GPT-6 Astra, plus an Ultrafast mode that runs 8x quicker behind a new $500 subscription tier. ChatGPT also grew Spaces, Pages, Slides, a meeting note-taker, and app-building plugins, turning it into a full office suite. The practical takeaway: basic AI wrappers and note-taking apps are now exposed, but deep industry expertise still isn't something OpenAI shipped.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 00:13

01 · Cold open

The claim that opens the video: OpenAI just killed a thousand startups in under an hour.

00:13 – 01:07

02 · Dev Day at a glance

The keynote ran about 50 minutes, shipped 20+ releases, and reaches ChatGPT's 1.2 billion weekly users.

01:07 – 02:57

03 · DOTS: the always-on agent

DOTS gets its own cloud computer and browser, connects to 4,000+ apps, watches for tasks like an unsent invoice, and stumbles through a live voice demo.

02:57 – 04:21

04 · GPT-6.1 Sol: cheaper, faster model

The new flagship-adjacent model costs about a third of GPT-6 Astra, claims to beat Claude 5.5 Opus on some benchmarks, and adds an 8x-faster Ultrafast setting.

04:21 – 05:27

05 · The pricing shakeup

Ultrafast is gated behind a new $500 plan, the $200 plan returns with less usage, and every existing user's limits reset to zero.

05:27 – 06:20

06 · ChatGPT becomes an office suite

Spaces, Pages, and Slides turn ChatGPT into a Notion/Google Slides competitor, plus a meeting note-taker that deletes its own audio.

06:20 – 07:15

07 · Plugins and sign-in with ChatGPT

Developers can build apps that live inside ChatGPT's chat panel, and users can pay for third-party AI features with their existing ChatGPT subscription.

07:15 – 08:31

08 · Which startups are actually in trouble

Basic wrappers and simple note-taking apps are exposed, but deep industry-specific expertise still isn't something OpenAI shipped.

08:31 – 08:51

09 · Outro

Points viewers to the full release list and a companion video on the GPT-6 Astra launch.

Atomic Insights

Lines worth screenshotting.

OpenAI's DevDay squeezed more than 20 product releases into a roughly 50-minute keynote.
DOTS, OpenAI's new always-on agent, gets its own cloud computer and browser, and can currently only be created once per account.
Every DOTS agent runs on GPT-6 Astra, OpenAI's most powerful model, and connects to more than 4,000 third-party apps.
GPT-6.1 Sol costs $2 per million input tokens and $10 per million output tokens, versus $10 and $50 for GPT-6 Astra, about a third of the price.
OpenAI claims GPT-6.1 Sol beats Claude 5.5 Opus on some benchmarks, though that comparison comes from OpenAI's own chart.
GPT-6.1 Sol currently only works inside ChatGPT's work mode and Codex, not the standard chat interface.
Ultrafast mode runs models roughly 8x quicker, up to 300 tokens per second, but it's gated behind a new $500-per-month plan.
OpenAI reopened its $200 plan alongside the new $500 tier, but with less usage included, pushing power users toward the pricier option.
ChatGPT is expanding into a full office suite: Spaces for shared team files, Pages for Notion-style docs, and Slides for presentations.
A new meeting note-taker records audio, deletes it once notes are generated, and can be triggered directly from Slack.
Developers can now build full plugin apps that live inside ChatGPT's chat panel and get recommended to its 1.2 billion weekly users.
Sign in with ChatGPT lets a user's existing subscription pay for AI usage inside third-party apps like Notion.
Simple AI wrappers and basic note-taking apps are the startups most exposed by DevDay, since OpenAI just built native versions of both.
Deep, industry-specific expertise wasn't something OpenAI shipped at DevDay, which is why it's argued to still be the strongest moat for founders.

Takeaway

Which startups DevDay actually threatens

WRAPPER RISK

OpenAI's DevDay didn't just ship features, it absorbed the exact jobs thin AI wrappers were selling, which means the surviving startups are the ones with real domain depth.

03DOTS: the always-on agent

An always-on agent like DOTS that watches for tasks and acts on them removes the reason to buy a single-purpose automation tool for the same job.
DOTS connecting to 4,000+ apps out of the box means integration breadth is no longer a differentiator you can charge a premium for.

04GPT-6.1 Sol: cheaper, faster model

GPT-6.1 Sol proves flagship-adjacent models keep getting cheaper fast, so pricing yourself against today's model costs is a short-lived strategy.
A model being gated to work mode and Codex only shows OpenAI staggers releases by product surface, not by user tier, worth watching before you assume a new model is universally available.

05The pricing shakeup

A new $500 tier alongside a scaled-back $200 tier signals OpenAI is segmenting power users by willingness to pay, not just usage volume.
Resetting everyone's usage limits to zero on launch day is a cheap, high-goodwill move that costs OpenAI little but buys a wave of renewed engagement.

06ChatGPT becomes an office suite

ChatGPT growing its own docs, slides, and meeting-notes tools means products that are just a thin interface over GPT-6 now compete directly with OpenAI itself.
A meeting note-taker that deletes its own audio after generating notes is a privacy detail worth copying in any product that touches sensitive recordings.

07Plugins and sign-in with ChatGPT

Letting developers build plugin apps that surface inside ChatGPT's own chat panel turns distribution into something that runs through the platform, not just your own marketing.
Sign-in-with-ChatGPT lets a user's existing subscription pay for your product's AI usage, which lowers the adoption barrier but also ties your economics to OpenAI's pricing.

08Which startups are actually in trouble

Simple AI wrappers and basic note-taking apps are the most exposed category, since OpenAI just built native, free-adjacent versions of both.
Deep, specific industry knowledge remains the one input OpenAI didn't ship at DevDay, and it's still what decides whether a tool gets adopted and paid for.

Glossary

Terms worth knowing.

DOTS: OpenAI's new always-on AI agent that runs on its own cloud computer and browser, works continuously in the background, and connects to thousands of third-party apps.
GPT-6 Astra: OpenAI's most powerful and most expensive model at the time of DevDay, priced at $10 per million input tokens and $50 per million output tokens.
GPT-6.1 Sol: A faster, cheaper OpenAI model released at DevDay, priced at roughly a third of GPT-6 Astra and currently limited to ChatGPT's work mode and Codex.
Ultrafast: A ChatGPT setting that runs generation roughly 8x faster, up to 300 tokens per second, available only on a new $500-per-month plan.
Cached input tokens: Repeated input tokens OpenAI charges at a steep discount versus fresh tokens, since the model doesn't have to reprocess them from scratch.
Sign in with ChatGPT: A login option that lets a user's existing ChatGPT subscription and credits pay for AI features inside other companies' apps.
Plugin extensions: Developer-built apps that live inside ChatGPT's own interface and get surfaced as recommendations while a user is chatting.

Resources

Things they pointed at.

02:55productMuse (Meta's personal assistant)

04:45productGrok Bot

06:06channelTechCrunch

08:05linkChristian Peverelli's free community

Quotables

Lines you could clip.

00:00

“Today, OpenAI killed more than a thousand startups in less than one hour.”

tight punchline hook, no setup needed→ TikTok hook↗ Tweet quote

03:50

“Almost as smart as Astra and way cheaper and quicker.”

one-line summary of the model pitch→ IG reel cold open↗ Tweet quote

07:25

“Definitely thousands of startups died today, my friends.”

restates the thesis mid-video, quotable on its own→ newsletter pull-quote↗ Tweet quote

08:05

“If you spent the last 10 years in a specific industry and you understand the ins and outs of it, you're literally the best positioned to actually build something and monetize it in this space.”

the actual advice underneath the clickbait framing→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

metaphor

Today, OpenAI killed more than a thousand startups in less than one hour. You heard that right. A couple hours ago at Dev Day, they gave ChatGPT its own computer, announced an assistant called DOTS that works 24 -7, and announced more than 20 different releases coming out today.

And so in this video, I'll be breaking down everything you need to know as a non -techie, and of course, in plain English. Let's go. It ran for about 50 minutes, and Sam Altman himself, the CEO of the company, said that it was the best Dev Day ever.

Over 20 different launches. Totally insane. Now, ChatGPT currently has 1 .2 billion users, meaning that anything that they release goes in front of a huge number of people.

That's why they released so many things at once. Now, they'll be rolling out the majority of those releases over the next coming days, and they've already made some announcements around that. I currently, as we speak, don't have access to all of these releases.

As soon as I do, I'll be building with, but probably the biggest announcements was DOTS. And if you don't remember anything else about today's presentation, you just need to know that DOTS released. So today, we're introducing DOTS.

It's probably a direct answer to Meta. who released their personal assistant called Muse. This pretty much allows ChatGPT to work without you having to be in front of your computer 24 -7.

Now, every dot or agent, at first you can only create one, basically runs on Astra 6, which is their most powerful model. And each one of these assistants gets its own computer in the cloud and also its own browser. As usual, you can connect this with more than 4 ,000 apps, all the ones you already use.

Email, Calendar, Slack, CRM. And then you can talk to it as a teammate. They even have their voice model attached to it, so you can literally just call it up and just have a chat.

And apparently it even watches for things that you will have to do and basically comes with those things to you, which if that is true, is pretty crazy. The example they gave live was basically someone who had forgot to create an invoice. And then the AI basically built the invoice for them, sent it to them for approval.

And once they had approved, they sent it off to the person who needed it. Now they're saying this rolls out to pro users, to enterprise users and to business users all today. And they also say that in the future, you'll be able to create multiple dots at the moment, you're limited to one.

And of course, not all of this is perfect. On stage, actually, there was this really uncomfortable moment where Holly, one of the presenters, actually gets stuck because voice doesn't work. Checking that now.

Still checking.

All right, I think maybe Dottie is having kind of a slow morning. And pretty much throughout this entire presentation, voice doesn't really work that well. Let's try again, live demos.

We might have to type this time, it's okay. But probably for a lot of people watching this video, we will already be a couple weeks or days away from the actual release, so you can expect to have a lot more stability. And so even though it's not perfect yet, it's a huge answer to Grokbot and Muse released by Meta.

Now the next up is probably the most exciting for me because it is the release of the new model. The name of this model is GPT -6 Soul, and because Astra's biggest criticism was to be too slow and also to be too expensive, they basically worked and improved those dramatically for this new release. Look at the price table.

Astra costs $10 per million tokens in and $50 out. Soul is $2 in and $10 out. Tokens are basically chunks of words.

but how powerful is this? We think you will really love this model that is smarter than Astra in some ways. They also showed a chart that showed that it beats Claude 5 .5 Opus.

They also mentioned it's about a third of the cost. Of course, that's their chart, so we're going to have to see and test it out for ourselves. Now, the catch is that this is currently just in ChatGPT work and codecs, so you won't be able to find it in the normal chat interface.

But yeah, almost as smart as Astra and way cheaper and quicker. That's a wild thing, but even cooler. Ultra Fast is now a new setting that you can actually use, meaning that you literally can have things done at the speed of light.

On stage, they showcased this by basically showing it build like a rocket sort of 3D design. And by the time like the first one was kind of in the middle of building the rocket, you were already seeing the other one just like take off. So it seems like it's way quicker on paper.

Again, we're gonna have to see all this in practice too. But yeah, Ultra Fast is a setting that makes it eight times quicker or about 300 tokens per second.

However, the catch here is that they created a new plan to get access to it, which is the $500 plan. So if you want it in your chat GPT, you're going to need to pay a premium for that. Now on the other side, they opened up the $200 plan again, but it has less usage in it.

So yeah, they're kind of trying to bring back the $200 plan, except it's a little bit less worth it. They're trying to push everyone to the $500 plan pretty clearly. Now the cool thing is that if you're already on a plan, you're going to stay on the exact usage, which is great.

They had like a button that was clicked and everyone got a reset. Three, two, one.

That's obviously for everyone in the world, not just people here at Dev Day. So now everyone, if you go back inside of your ChatGPT, you should have all your tokens back, all the limits set again to zero, and you should be able to do a lot more work, at least today or over the next couple of days. Now, another super interesting announcement, for me at least, was that they're kind of turning ChatGPT into an online suite.

First, it was basically Spaces, which is a place you can work with all of your colleagues and also your DOT, which is the agent we talked about earlier. have all access to the same files. Inside of that, they have pages, which is kind of like a notion where you can actually tag people, you can work within specific documents, create new ones, leave comments for other people, you can generate charts.

It's pretty much notion. And they also announced that they're going to be releasing slides, which obviously is pretty much like being able to create all the slide decks, so like Google Slides. that's only going to come in the next couple of weeks they also announced a meeting note taker that kind of disappears and deletes the audio once it's done but puts it all in notes a cool feature where you can tag it inside of slack so if you want to basically ask for things to happen from within your slack you can do that as well and techcrunch called it their own office so think of it like office 365 basically now they're coming after the full workstation that ai works within and you can get access to these on pro on business and on enterprise at the moment but actually this next release is something I'm also incredibly excited about.

Well, now developers can build full apps that live within ChatGPT. They're called plugin extensions, and basically, your app gets a spot in the panel that they're already chatting with. So these basically come up as recommendations once they're actually chatting with AI.

I need to do this. Oh, they recommend a couple plugins. Yours could be one of those.

They actually showcased what you could do with these as well, which was quite cool. So the whole pitch was, hey, you can just build it and then you're going to get access to our customers, right? So 1 .2 billion people, you're going to get recommended to some of those people as they're in the mode of chatting.

Then another cool, interesting development was sign in with ChatGPT. It's basically a way for people to bring their own keys to your software. So now if you have a subscription for ChatGPT, for example, and you want to log into Notion with ChatGPT, you can use your already existing credits inside your subscription to do things inside of Notion.

Now, Sam himself admitted that there were some mixed results with this earliest thing, but he really believes in the future of this. Now, one thing that I noticed about this release is that a lot of people are actually scared about all these releases, and definitely thousands of startups died today, my friends. Specifically, the companies that have very basic wrappers are in trouble, right?

So if you only have a very... simple note -taking app for example you're in trouble we also saw that the entire sort of suite of where you run things to run your ai is also in trouble right so they're actually also taking stabs at really big players out there like google but the thing i found quite interesting is that they really didn't go too too deep in any specific vertical here right so still having that industry specific knowledge knowing how to actually build something that drives value for that specific customer and then figuring out how to get paid for it are still really important parts of the entrepreneurship the opportunities that are available are simply going to change but if you spent the last 10 years in a specific industry and you understand the ins and outs of it you're literally the best positioned to actually build something and monetize it in this space leveraging the power of these tools and these were just a couple of more than 20 releases that chad gpt announced today but if you want the full list you can definitely download it for free in my free community click below in the description and you'll find it

And although everything they announced today is incredibly powerful, if you missed the release of Astra, which is their most powerful model, yes, it's more expensive, but it's still the leading model for OpenAI, then definitely feel free to check out this video because I tell you everything about the announcement, how it works, and show you the power of it.

Hope you enjoyed this video. I'll see you in the next one. Let's go.

The Hook

The bait, then the rug-pull.

OpenAI's DevDay ran less than an hour and shipped more than 20 releases at once, from an always-on agent called DOTS to a full in-app office suite. This breakdown translates the keynote into plain English and flags exactly which kinds of startups just got squeezed.

CTA Breakdown

How they asked for the click.

VERBAL ASK

08:05link

“download the full list for free in my free community, click below in the description”

soft, single mention near the end, tied to the value of the full 20+ release list rather than a hard sales pitch

FROM THE DESCRIPTION

PRIMARY CTAWhere the creator wants you to go next.

✅ Free Resources ↗

OTHER LINKSAlso linked in the description.

Storyboard

Visual structure at a glance.

open

hookopen00:00

DOTS

valueDOTS01:07

GPT-6.1 Sol

valueGPT-6.1 Sol02:57

startup fallout

valuestartup fallout07:15

CTA

ctaCTA08:05

Frame Gallery

Visual moments.

open

Frame at 00:07 from OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

Frame at 00:14 from OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

Frame at 00:25 from OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

Frame at 00:30 from OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

Frame at 00:35 from OpenAI Just Killed 1,000 Startups: Every DevDay Announcement Explained

One-click upgrade to your Google

Get more breakdowns in your search results

Add Modern Creator as a preferred source and Google shows you more of our breakdowns in Search, Top Stories, and AI Overviews. It only changes what you see, and you can undo it in your Google settings anytime.

Add to Preferred SourcesOpens your Google source preferences with us pre-loaded. Tick the box and you're done.

Watch next

More from this channel + related breakdowns.

17:45

Christian Peverelli · Demo

I Built 3 Projects in GPT-6 Astra vs Fable 5.1: Which Is Better?

One creator ran the same three builds through OpenAI's GPT-6 Astra and Anthropic's Fable 5.1, then found the real gap in what happened after he stopped typing.

September 8th

20:32

Alex Finn · Reaction

OpenAI Dev Day: Every Announcement Explained

A creator back from OpenAI Dev Day walks through every launch, from a unified Dots agent to a $500 ChatGPT plan, with early hands-on impressions of each.

September 30th

32:33

Stacked Podcast · Interview

How Claude Opus Helped Researchers Breach OpenAI in 3 Hours

A sanctioned pen-testing team used Claude Opus to crack OpenAI's private code repo for a $6,500 bounty, then the hosts spend the rest of the episode on Claude Projects, Astra for Law, and Anthropic's claim that Claude now runs a quarter of its own AI research.

September 18th

17:58

Corey McClain · Talking Head

ChatGPT Voice Just Killed the Keyboard

OpenAI rebuilt voice mode around GPT-5.6 and GPT-6 Astra, gave paying users up to unlimited daily hours, and made the case that talking may replace typing as the default way to use AI.

September 10th

19:32

Matt Wolfe · Review

GPT-6 Astra Is Finally Here (And It's Really Good)

A first-look review of OpenAI's GPT-6 Astra, run through published benchmarks, a set of repeatable creative tests, and a computer-use experiment where the model built and animated its own 3D game world.

September 3rd

02:43

OpenAI · Demo

Introducing GPT-6 Astra

OpenAI's product-reveal film for its new ambient assistant, staged as one continuous demo running six unrelated real-world tasks at once.

September 3rd