Was GPT-5 Underwhelming? OpenAI Co-founder Leaves, Figure02 Arrives, Character.AI Gutted, GPT-5 2025

Did GPT-5 just get delayed to 2025? Brockman takes leave, Schulman just leaves and OpenAI co-founders are abandoning ship. Plus Figure02, Musk lawsuit, Character.AI hollowed out, new Gemini 1.5 Pro, Dev Day hints and more. The race narrows to just Meta, Anthropic, Google and OpenAI. I’ll cover what this all says about the eventual release of GPT-5 and the race to ever smarter models…

[inspectorcrud]

These humanoid robots look superficially good, I need to see them doing something more complicated than pick and place or loading a coffee machine. I need to see them navigate an unfamiliar house, climb the stairs with a tray of food it has prepared.

[DaveShap]

My spidey sense was just tingling. Time for another AI Explained video.

[HNi]

I love how structurally critical you are of AI news and studies. This is great journalism. You're one of my main sources for AI news. Great quality.

Tiny feedback, I believe many of us would highly appreciate it if you add chapters (sections) to your videos. It makes it easier to reference a certain part or skip to an anticipated part. Especially since you talk about different topics.

[hypnonaut]

Still the best and most reliable A.I. channel on YouTube

[alphaorg]

"Lex posted 36 hour podcast with Ilya 10 hours ago, and I 4x speed through. Here are the highlights." -Philip (Neuralink upgraded), AI Explained in 2026

[DerekSmit]

Maybe OpenAI has a problem where they can't just "steal" data anymore. Meta, Google and even Twitter have a lot of user data. OpenAI isn't really a company with it's own data. In the past they would just scrape the entire internet, but that is getting harder and harder, to a point where maybe they can't advance anymore without more and better training data.

[dannyquiroz]

isn’t this what we wanted? a competitive ai landscape…win?

[augustuslxiii]

0:40 Imagine saying "Yeah, I'm going to take the rest of the year off" in August, at your job.

Must be nice.

[dan-cjrr]

"The mission is far from complete; we still have a safe AGI to build." so they already have an unsafe AGI ?

[mm-rjvo]

"Breaking news, richest man on the planet accuses less-rich man of greed, onlookers ask "What else is new?". More at 5."

[bosco]

Earlier this year you quoted an interview with someone who said we wouldn't see any huge advances in ai this year, more likely we would see the hype die down and companies polish their ai services / products and work out how to actually sell it to enterprise, with the next big leaps coming next year. Seems spot on to me.

[JohnSmithAB]

OpenAI is still pulling further ahead in the breathless hype race. "More advances will soon follow and will usher in a decisive period in the story of human society"!? WTF dude calm down and ship something.

[nathanlannan]

chatbot arena often comes down to which has better markdown formatting - in this regard the gpt4 suite is very good

[sbowesuk]

When we add up all the product delays and irreplaceable talent leaving in droves, it's abundantly clear that OpenAI is dysfunctional. Maybe it would have been best if Sam had stayed fired...

[trucid]

Sonnet 3.5 refuses less than Opus 3, so let's hope the trend continues and Anthropic works on reducing the refusals further.

[dishcleaner]

Having the best AI model be open source is probably a better scenario than the dude having people scan their retina for a Ponzi having the reigns.

[JD-jlyy]

I hope Altman doesn't gets his way. We don't need machiavellianism incarnate at the steering wheel of AI

[Scbas-YT]

counting to 50 in one breath is quite the feat... for humans! LOL

[kiwihame]

I swear Chat-GPT is getting more dumb and less useful. Even Perplexity is hallucinating like crazy now.

[Caldaron]

they have to hype for agi because of funding. even though its too far away and needs a couple of major breakthroughs