The skinny on what's happening in AI - straight from the previous live session:
Five days into August and people are already losing their minds asking "where's my GPT-5?" after Sam Altman mentioned an August timeline.
OpenAI generally releases on Thursdays, so earliest would be August 7th, but Sam's recent tweet warning of "probable hiccups and capacity crunches" suggests they're prepping servers by potentially reducing current model capacity.
we have a ton of stuff to launch over the next couple of months--new models, products, features, and more.
please bear with us through some probable hiccups and capacity crunches. although it may be slightly choppy, we think you'll really love what we've created for you!
— Sam Altman (@sama)
5:22 PM • Aug 2, 2025
Anecdotally people are having ChatGPT “issues” at the moment too. Which is what always happens as server capacity gets shifted around in prep for a new model launch.
Kyle's take: We've waited months - we can wait a bit more. The entitlement is mental. Sam's basically doing damage control because every major model release absolutely demolishes their servers, and people will still moan when it happens.
The real question: will GPT-5 feel meaningfully better to normal users, or have we hit diminishing returns where improvements are too subtle for most people to notice? We’ll know soon!
Source: TechRadar coverage
Elon's launched Grok Imagine with a literal "spicy" button for creating NSFW video content. While other AI companies shy away from adult content, Musk is sprinting towards it - complete with Ani, their entirely not-safe-for-work AI avatar. This isn't an accident or oversight; they've built a spicy button specifically for soft-core content.
Kyle's take: Brilliant business move, questionable ethics. Musk knows there's massive demand for this stuff - and I'd be willing to bet 50% of usage is in spicy mode.
While OpenAI, Google, and Anthropic play it safe, Grok's claiming the market everyone else won't touch. They're positioning as the "free speech" AI whilst others get labeled as censored corporate tools.
Source: Grok Imagine launch / Verge Coverage
Sam Altman tweeted we're entering the "fast fashion era of SaaS" where vibe coding tools let anyone build basic software in hours.
Why pay DocuSign $75/month when you can build your own document signing tool? You are basically supporting their bloated payroll of 7000 (!!!) employees.
Kyle's take: This is spot on but brutal for legacy SaaS.
The switching costs and "nobody gets fired for buying IBM" mentality will keep big enterprises locked in, but solopreneurs and startups? They'll build everything themselves or use new more cost effective entrants’ offerings.
The word "fast fashion" is interesting though - suggests cheap, disposable, low-quality AI slop. Which is probably accurate for most vibe-coded solutions but it’s not in Altman’s benefit to point this out. I think he’s trolling a bit for engagement - he’s pretty damn good at that!
Source: Sam Altman tweet
Y Combinator partner says the "hardcore move" for AI automation founders is to actually get the job you want to automate first. Learn the pain points, understand the workflow, then build the solution.
They call getting a job "going undercover" - hilarious Silicon Valley speak for having actual work experience.
Kyle's take: This is brilliant advice but misses an obvious point - if you've worked in an industry for 10 years, you already have this intel. You don’t need to be a 20 year old software engineer “going undercover”.
You already know exactly what's broken, what's annoying, what takes forever. Skip the "undercover" nonsense and build solutions for problems you've personally felt. Every industry expert watching this is better positioned than some 19-year-old Stanford dropout pretending to be a payroll clerk.
It’s just up to you to seize the opportunity.
Source: Y Combinator discussion
Meta's leaked "super intelligence team" shows half the members are Chinese nationals. Not terribly surprising consider how good the education system is!
Mark's spending hundreds of millions to poach talent from Apple, OpenAI, and Chinese companies, but here's the (potential!) problem: when their contracts end, they'll likely return to China with everything they've learned about AGI development.
Kyle's take: This is the tech equivalent of teaching someone to fish then being surprised when they start their own fishing company. Who swiftly buys up your old fishing spot and kicks you out.
We've seen this playbook before - Japan became motorbike kings after the West outsourced manufacturing to them, China dominates semiconductors (and…everything manufactured!) after the West sent their best engineers there.
It’s a common pattern. And it’s about to happen again.
Mark's thinking short-term sprint to AGI, but he's also training China's next generation of AI leaders.
Personally I don’t have a problem with this - I’m pretty neutral here. But I can imagine this becoming contentious down the line. Zuck will be safely tucked up in his super-bunker by then though!
Source: Meta superintelligence team leak
Member Question from Leon: "Why would you need GPT-5? The current models are really good already."
Kyle's response: Spot on question. For 99% of users doing brainstorming and basic tasks, GPT-4O and Claude Sonnet/Opus are already brilliant.
The real value might be that older models get cheaper, creating a waterfall effect where today's premium AI becomes tomorrow's commodity pricing. Most people don't need the cutting edge - they need the previous generation at accessible prices.
But: maybe we’ll see something amazing with GPT-5! We’ll know soon!
This question was discussed at [04:00] during the live session.
Want the full unfiltered discussion? Join me tomorrow for the daily AI news live stream where we dig into the stories and you can ask questions directly.
Make us a Preferred Source on Google and catch more of our coverage in your feeds.
AI with Kyle Daily Update 172 Today in AI: Beginners Guide to Claude Code Kyle Balmer April 17, 2026 https://youtu.be/RG04qMlZz_s You've seen the viral…
