- AI with Kyle
- Posts
- AI with Kyle Daily Update 024
AI with Kyle Daily Update 024
Today in AI: OpenAI Drops Bombshell Open Source Models + GPT-5 Coming Tomorrow?
Kyle Balmer
August 07, 2025
The skinny on what's happening in AI - straight from the previous live session:
Highlights
🚀 OpenAI Just Entered the Open Source War
OpenAI just released their two gpt-oss models - two very solid open source offerings Two versions: 20B (runs on phones, apparently but I doubt it….) and 120B (needs proper hardware but theoretically could be run at home).
These aren't rubbish old models either - they're basically as good as o3 and 4o-mini, which until now has been their frontier models. But much cheaper.
gpt-oss is a big deal; it is a state-of-the-art open-weights reasoning model, with strong real-world performance comparable to o4-mini, that you can run locally on your own computer (or phone with the smaller size). We believe this is the best and most usable open model in the
— Sam Altman (@sama)
5:07 PM • Aug 5, 2025
Kyle's take: This is massive for three reasons. First, it completely changes the open source landscape just a week after Zuckerberg pulled Meta out of that race.
Second, Sam Altman's timing here aligns perfectly with Trump's AI action plan about keeping America competitive against China. Brownie points. Altman even declared this to is make sure US AI is first in the world.
Third, for solopreneurs, this could drive down your inference costs significantly if you're building AI applications. Though let's be honest - most people will still use the API because running these locally is bloody slow unless you have some serious GPU power.
Source: OpenAI announcement
⚡ GPT-5 Likely Dropping Tomorrow?
Sam Altman's been doing his usual Twitter teasing, saying "something big but small today" (the open source models) followed by "a big upgrade later this week." OpenAI historically loves Thursday releases to own the weekend news cycle.
My money is on Thursday 7th as release day.
Kyle's take: I've cleared my evening tomorrow to play with it because I'm 99% certain we're getting GPT-5 on Thursday US time. The open source release feels like they're clearing the deck - basically making their current models obsolete before the big reveal! Classic Apple playbook, but for AI models. Power play!
😬 Grok's "Spicy" Mode Creates Taylor Swift Deep Fakes
Elon's Grok just launched "Imagine" mode that can animate photos into 5-7 second clips. It includes a "spicy" button that immediately generates not-safe-for-work content, including celebrity deep fakes. Journalists tested it and got topless Taylor Swift videos without even asking for explicit content.
This is NOT Tay Tay
Kyle's take: On one hand, Elon's being realistic about what people actually want to use AI for - the adult market is massive. Let’s not kid ourselves.
On the other hand, he's handing governments exactly the ammunition they need to regulate the hell out of all of us. The UK, US and Australian governements are frothing at the mouth over age checks right now and Elon is feeding the beast.
While we're trying to prove the internet can self-regulate, he's out there making celebrity deep fake generators. Not helping our cause here….
Source: TechCrunch coverage
📊 Claude Gets 2% Better (Yay…but yawn)
Anthropic released Claude Opus 4.1 with a whopping 2% improvement on software engineering benchmarks. From 72.5% to 74.5% accuracy.
Kyle's take: I love Claude - it's my primary model - but this is exactly why nobody talks about Anthropic outside AI circles.
While Sam Altman's dropping open source bombshells and teasing GPT-5, Claude gets a 2% bump and calls it a day.
The Google Trends data doesn't lie - ChatGPT dominates public consciousness whilst Claude barely registers. Quality matters, but so does marketing, and Anthropic's rubbish at the latter.
ChatGPT IS AI as far as the public is concerned:
Source: Anthropic announcement
Member Question from James: "What AI jobs involve no coding experience in the UK? Can you see them increasing?"
Kyle's response: Brilliant question. Look for roles that are non-algorithmic (different every day), physical (require being there), and empathy-based (human connection).
Think care work - nursing homes, palliative care, home assistance. The aging population means these jobs will boom, and they're AI-proof because they need real human empathy and physical presence.
Also consider the trades - plumbing, electrical work. Until robotics catches up and becomes cheap enough, these are solid. I've got a mate who has a 9-5 day job and is a plumber. The plumbing pays great because there's a shortage of good tradespeople. Not sexy, but everyone needs their toilet fixed! AI won’t be changing that any time soon.
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.