Hang tight while we fetch the video data and transcripts. This only takes a moment.
Connecting to YouTube player…
Fetching transcript data…
We’ll display the transcript, summary, and all view options as soon as everything loads.
Next steps
Loading transcript tools…
OpenAI was dead… Then GPT-5.2 dropped | Fireship | YouTubeToText
YouTube Transcript: OpenAI was dead… Then GPT-5.2 dropped
Skip watching entire videos - get the full transcript, search for keywords, and copy with one click.
Share:
Video Transcript
Video Summary
Summary
Core Theme
OpenAI has responded to Google's Gemini dominance with the release of GPT 5.2, showcasing significant advancements in efficiency and reasoning capabilities, particularly on the ARC AGI benchmark, while also raising concerns about the ethical implications of AI-generated content and insider trading in prediction markets.
Mind Map
Click to expand
Click to explore the full interactive mind map • Zoom, pan, and navigate
Last week, Sam Alman declared a code red
threat level midnight after OpenAI's
lead in the AI race was shattered by the
unexpected dominance of Google's Gemini
3. Just about everybody was ready to
ride off OpenAI as the Netscape of the
2020s. That is until yesterday when they
dropped their answer to Gemini GPT 5.2,
a model that once again moves the AI
hype wheel back in favor of open AI.
It's dominating all the Trust Me Bro
benchmarks and even beats Claude Opus
4.5 on software engineering and
reasoning. That the real flex though is
its rise to the top of the ARC AGI
benchmark. The ARC prize just verified a
390x efficiency improvement in one year
from the 03 model to 5.2. That's not a
typo. That's a model that's 390 times
more efficient. In today's video, we'll
find out if we finally reach the edge of
the AGI threshold or if this is just
more smoke and mirrors to keep the AI
hype train going into 2026. It is
December 12th, 2025 and you're watching
the code report. Artificial intelligence
has already ruined my Christmas this year.
year.
>> It's the most terrible time of the year.
Thanks to this nightmarish commercial
produced by the artificial food
generation company McDonald's, the
creators of this piece of crap you're
watching tried to act like they prompt
engineered a real work of art, but it
was so universally hated, McDonald's was
forced to pull it from the airwaves.
Unfortunately, this AI slop content is
only going to get worse. Because OpenAI
just inked a $1 billion deal with Disney
to allow their iconic characters to
appear in AI generated photos and
videos. And that's huge because it means
anybody can now generate their own
custom Star Wars or Toy Story movie and
will be forced to use OpenAI's tech to
do so. Very concerning. But speaking of
concerning, prediction markets like Poly
Market and Kelshi that somehow predicted
that GPT 5.2 would be released
yesterday. And by predicted, I mean
OpenAI employees and other insiders
found a new infinite money glitch
because insider trading exists in a gray
area in these markets. An obvious
insider at Google made a million bucks
this month. And in many cases, it's
insider trading that makes these
prediction markets so accurate. Very
concerning. But now, let's get back to
GPT 5.2. And the thing everyone is
talking about, its performance on the
ARC AGI benchmark. But what even is ARC?
It stands for abstraction and reasoning
corpus and is designed to test whether a
model can solve novel unique problems
it's never seen before. Problems that
require pure reasoning instead of
memorization. The problems are
intentionally weird low data puzzles
where brute force pattern mashing fails.
Regular humans can usually solve them
after a few examples. What's weird
though is most AI models completely face
plant. The important takeaway is that a
model that scores well on ARC has the
ability to generalize instead of just
acting like an autocomplete on steroids.
And that's why OpenAI flexing on this
chart is a much bigger deal than many of
the other Trust Me Bro benchmarks. But
for the average LLM user, it's becoming
harder and harder to evaluate each new
release. Like ChatGpt 5.2 2 is also
supposedly much better at coding than
before and it's supposed to have far
fewer hallucinations, but I'm not sure I
can even tell a difference. I'm happily
using it to generate spelt 5 code with
an MCP server, but what I'm more
concerned about is deploying my code
somewhere reliable, like Railway, the
sponsor of today's video. It's a cloud
platform that lets you instantly host
productionready deployments and manage
your entire infrastructure stack in one
place. is so instead of fighting with
mystery YAML files and 25 different
dashboards, you can spin up isolated
environments in one click that will
scale automatically as needed. And
unlike other platforms that shall not be
named, Railway only charges you for the
resources you actually use, not what you
provision, which saves over 65% on cloud
costs. And developers love it because it
gives you 50% faster build times while
letting you spin up any service you need
shockingly fast. They even have 1,800
different templates that let you deploy
any app or database with a single click.
Sign up for free today at the link below
and you'll get $20 in credits when you
upgrade. This has been the Code Report.
Thanks for watching and I will see you
Click on any text or timestamp to jump to that moment in the video
Share:
Most transcripts ready in under 5 seconds
One-Click Copy125+ LanguagesSearch ContentJump to Timestamps
Paste YouTube URL
Enter any YouTube video link to get the full transcript
Transcript Extraction Form
Most transcripts ready in under 5 seconds
Get Our Chrome Extension
Get transcripts instantly without leaving YouTube. Install our Chrome extension for one-click access to any video's transcript directly on the watch page.
