• Everyday AI
  • Posts
  • Ep 472: OpenAI’s new GPT-4.5: What’s new and who can benefit the most

Ep 472: OpenAI’s new GPT-4.5: What’s new and who can benefit the most

Our review of GPT-4.5, Google's scary office commitment, and Meta's trying to compete with OpenAI in a new way.

Author

Everyday AI
February 28, 2025

Outsmart The Future

Today in Everyday AI
9 minute read

🎙 Daily Podcast Episode: OpenAI just released its newest model, GPT-4.5. And…. it’s weird. In a good way. Give it a listen.

🕵️‍♂️ Fresh Finds: Google's new AI spreadsheet flex, GPT-4.5 getting mixed reviews, Sora launches in EU and UK. Read on for Fresh Finds.

🗞 Byte Sized Daily AI News: Google cofounder wants workers to put in 60-hour workweeks to replaces themselves, Meta's trying to compete with OpenAI in a new space, and people are up in arms about Donald Trump's latest use of AI and more, read on for Byte Sized News.

🚀 AI In 5: Have you tried these 3 Google Gemini live features? See it here

🧠 Learn & Leveraging AI: GPT-4.5 has people all confused. We break it down and tell you how to take advantage of everything OpenAI’s new powerhouse model has to offer. Keep reading for that!

↩️ Don’t miss out: GPT-4.5 released, Amazon unveils quantum chip, NVIDIA revenue soars, Microsoft calls for removal of AI chip restrictions and more!  Check it here!

 Inside Multi-Agent AI: Rethinking Enterprise Decisions 💡

 How big is ChatGPT's new GPT-4.5 model? 🥵 Welp. They ran outta GPUs.

↳ 10X compute power. ↳ Sharper coding skills. ↳ Vision that finally gets it. ↳ And writing with actual nuance instead of AI word salad.

Is this mid-step evolution worth the API price jump?

Who wins biggest from these upgrades?

Join us as we break down what matters in 4.5—from its creative intuition to its technical precision—and why that 2023 knowledge cutoff might not be the dealbreaker you think.

Also on the pod today:

•Deep Dive into GPT 4.5 🤿
• Who Can Benefit from GPT 4.5 🤷‍♂️
• 4.5’s Technical and Strategic Impact 🤓

It’ll be worth your 52 minutes:

Listen on our site:

Click to listen

Subscribe and listen on your favorite podcast platform

Listen on:

Here’s our favorite AI finds from across the web:

New AI Tool Spotlight – Promptimize ups your prompting game, Mesh is a new AI bookkeeper for startups (bye Quickbooks?) and Pig is like OpenAI’s Operator, but for PC.

ChatGPT — We already gave you our thoughts on GPT-4.5, but the internet has super mixed reviews.

AI Legislation – California’s SB 53 aims to boost AI innovation with low-cost computing for startups while safeguarding whistleblower.

Google’s AI – Google Sheets lovers rejoice: Google’s AI now allow you to take your data visualization game to the next level.

LLM Rankings – Claude’s new 3.7 model got the LM Arena ranking treatment. 

AI Safety — Researchers found that AI wants to rule over humans. Sounds…. fun?

AI Video — After like forever and a half, OpenAI’s Sora model launched in EU and UK.

AI in Hospitality – Las Vegas (of course) will be the home to the world’s first AI-powered hotel.

1. Sergey Brin Urges Google Engineers Back to Office in AI Race 🏃‍♂️

Google co-founder Sergey Brin is calling on engineers to return to the office five days a week, pushing for a "60-hour sweet spot" of productivity to help the company win the AI arms race, according to The New York Times.

Following OpenAI’s ChatGPT success, Brin—worth $144 billion—has reemerged at Google’s Mountain View HQ, urging employees working on its Gemini AI models to accelerate their efforts.

While Brin promotes using Google’s AI tools to turbocharge coding efficiency, critics highlight the irony of AI potentially replacing engineers or creating more workload for them.

2. 1,000 Scientists Test AI Models in National Labs Experiment 🔬

Nearly 1,000 scientists from nine U.S. National Laboratories gathered Friday for a groundbreaking AI testing event hosted by OpenAI and the Department of Energy, according to The Hill.

Dubbed the “1,000 Scientist AI Jam Session,” the event tested cutting-edge models like OpenAI’s o3-mini and Anthropic’s Claude 3.7 Sonnet to explore how AI can accelerate scientific discovery and national security efforts. Anthropic’s new hybrid reasoning model, touted for its ability to switch between quick answers and deep thinking, was put through real-world problem-solving challenges like experiment planning and analysis.

3. Meta Eyes Standalone AI App to Rival OpenAI and Alphabet 📲

Meta is reportedly developing a standalone app for its Meta AI chatbot, aiming to break free from its social media platforms like Facebook and Instagram, according to CNBC.

This move could help Meta capture millions of users outside its ecosystem, as Zuckerberg pushes to make Meta AI the go-to assistant for over a billion people. The app may also introduce paid plans, signaling Meta’s ambition to monetize its AI offering and compete with industry giants like OpenAI’s ChatGPT and Google’s Bard.

4. Amazon’s Alexa+ Gets a Claude Boost Amid AI Race 🧠

Amazon is reportedly relying on Anthropic’s Claude language model to power the most advanced features of its new Alexa+ service, according to sources cited by CNBC, despite Amazon denying the claim.

The revamped Alexa, launching soon at $19.99 per month, can now perform complex tasks like booking rides and making reservations—capabilities that showcase Claude’s “intellectual heft,” insiders say. While Amazon emphasizes its in-house Nova AI model and Bedrock platform, sources note that Claude has been key to Alexa’s leap forward as Amazon continues its $8 billion partnership with Anthropic.

About time Alexa finally gets smart. Bless up.

5. Trump Shares Satirical AI Video, Creators Cry Foul 😭

A satirical AI-generated video depicting a bizarre "Trump Gaza resort" has sparked controversy after being shared by Donald Trump on social media, according to NBC News.

The creators, Solo Avital and Ariel Vromen of EyeMix Visuals, say the video, which features caricatures like bearded bikini-clad dancers and a golden Trump statue, was meant as satire—not propaganda. They claim they never intended for it to go viral and were shocked when Trump posted it without context or credit, calling his use of their work "stolen content."

6. Global AI Child Abuse Crackdown Unveiled 🦺

A Europol-led operation has arrested 25 individuals worldwide for distributing AI-generated child sexual abuse material, highlighting the alarming growth of such content online.

Dubbed "Operation Cumberland," this marks one of the first cases targeting AI-generated exploitation, exposing gaps in national laws to handle these crimes. While the U.S. did not participate, this comes as lawmakers scramble to address AI-manipulated "deepfake" imagery, with the Senate recently passing the "TAKE IT DOWN Act" to combat non-consensual intimate content.

3 Must-See Features Inside Google Gemini Live

Have you used these?

If you haven’t used Google Gemini Live, we give you three convincing reasons why you may want to start.

🦾How You Can Leverage:

GPT-4.5 isn’t for us.

Let’s be honest, shorties.

Most of us are power users, orchestrating multiple LLM outputs like a soccer mom policing chocolate-wasted kids in the back of a van.

Even though it’s been barely a day since OpenAI rolled out GPT-4.5, we see power AI users throwing sad-faced emojis all over the place.

But let’s be clear.

GPT-4.5 isn’t built for us.

This one’s for the other ninety-nine percent of the world that barely touches a single LLM, let alone five.

With this release, OpenAI is marching away from chart-topping specs and chasing emotional intelligence instead. They’re banking on the millions who will swallow these updates whole, not on the one percent barking for trillion-parameter wonders.

And that’s just fine.

(But still…. BARK BARK MORE FEATURES! Lolz) 

We think OpenAI is focusing on the hundreds of millions of users who will gulp up these EQ updates, and not the 1% of us demanding a trillion features and perfect benchmarks. 

And that’s OK.

Hot off the presses of OpenAI’s newest release, we gave GPT-4.5 an in-depth run through. 

But if you did, here’s the 1-2-3 of what you need to know y’all. 👇

1 – GPT-4.5 Won’t Win Benchmark Trophies 🏆

GPT-4.5 doesn’t care about setting test score records. It’s built to feel.

We combed through the conversation and saw the same trend repeated: GPT-4.5 is more approachable, more empathetic, and ironically, more relevant to everyday people. 

This shift is leaving AI super-nerds disappointed. But it’s the rest of the planet that stands to gain.

Why didn’t we see this coming?

Because we expected brute force. We expected bigger data centers and trillion-parameter extravaganzas. (But yeah, that’s still happening too. 

But GPT-4.5 took a side road, focusing on conversation flow and emotional nuance.

Look at the cost. It’s 30 times pricier than GPT-4o on the API side, making it a ridiculous choice if your devs are trying to process gazillions of tokens daily. 

Yet it’s perfect for smaller organizations needing a human feel.

Maybe you only serve a few thousand support requests per week. Maybe you’re a niche coaching practice. GPT-4.5 offers hyper-human interactions that don’t feel so mechanical.

Try This

Use GPT-4.5 for client-facing interactions that hinge on trust. Onboard it for high-ticket sales or post-sale follow-ups where empathy matters more than volume.

Write your prompts in a warm, personal voice. Gauge responses through direct feedback or short surveys. Ask clients if they sense a difference in tone and accuracy. If they say it feels like chatting with a knowledgeable friend, you’re on the right track.

Measure how quickly issues get resolved compared to your usual AI model. If resolution times shrink and user satisfaction grows, congratulations. You’ve tapped the real power of GPT-4.5.

2 – This Model Isn’t For the 1% of Power Users 💪

One point is crystal clear: we, the AI addicts who juggle multiple chatbots daily, aren’t the crowd GPT-4.5 caters to.

Yeah, yeah, yeah. Right now, it’s only available to the power users on the $200/month Pro plan. But, it’ll be rolled out to $20/month Plus users soon. 

(And we expect free users will eventually get limited 4.5 queries.) 

We think GPT-4.5 is actually targeting that vast market of everyday humans who just need a comfortable, accurate sidekick.

A creative brainstorming partner. A counter-point strategist that has ALL the tools in the bag. 

Yeah, it’s slower. It’s also not aiming for the “fastest code generator” trophy. The real story is how GPT-4.5 connects with casual users who’ve been wary of AI. Think small businesses, local shops, or new adopters in larger enterprises.

Or your mom. 

We picked up on something else. GPT-4.5 might anchor an upcoming line of smaller, domain-specific models that rely on its emotional intelligence baseline. 

Why? Because if OpenAI nails the vibe piece, domain experts can tune specialized spin-offs.

This approach signals a massive pivot. Instead of brute force, we get curated experiences. 

The data from GPT-4.5’s chats will likely train new models to handle, say, HR tasks or healthcare triage with higher empathy.

For some companies, that’s the key to AI adoption. They don’t want a machine spouting nonsense or sounding cold. 

They want a sense of relationship. GPT-4.5 cracks that door wide open.

Try This

Pinpoint a department in your company where employees are hesitant about AI. Let GPT-4.5 prove it can “talk human.”

Draft a pilot project. Maybe it’s HR responding to employee questions about benefits. Keep it small and personal. Let GPT-4.5 handle a limited set of queries.

Track how often human intervention is needed. Compare overall sentiment before and after the pilot. If confusion or frustration drops, and employees say the AI feels more approachable, you’re seeing GPT-4.5’s biggest strength in action.

3 – The Future is Relatability and Reliability 🤝

GPT-4.5 is likely the final stepping stone before OpenAI merges its GPT architecture with more advanced reasoning models in the o-series. 

Users have complained about hallucinations. GPT-4.5 reportedly cuts them drastically. Early tests show it’s not immune to lying, but it’s better at filtering out nonsense.

For those who’ve avoided AI due to inaccuracy fears, that’s a big deal.

There’s something more profound. This is a reliability upgrade. Fewer random facts. More consistent logic. That opens doors in fields like customer service, where confusion can cost you a sale.

Reliability is also about user trust. People don’t mind if the AI isn’t super-genius, as long as it feels honest and friendly.

GPT-4.5 leans into that sweet spot, leaving power users to grumble but everyday adopters to cheer.

Why does this matter now?

Because the next wave of AI breakthroughs won’t hinge on raw IQ. They’ll hinge on building trust at scale with EQ. 

And vibes, apparently. 

GPT-4.5 is a sneak peek into that future. And it’s probably unstoppable if it nails these twin pillars of relatability and reliability.

Try This

Test GPT-4.5 in a critical consumer-facing role. Let’s say you run a subscription service with frequent troubleshooting requests.

Feed real user scenarios into GPT-4.5. Ask for step-by-step solutions. Look for any hints of hallucination or confusion. Then measure how consistently it delivers correct answers.

Finally, gather user feedback. Do they trust the AI as much as your human reps? If trust and satisfaction rates climb, keep GPT-4.5 in the loop. That’s where it outshines most other models today.

⌚

Numbers to watch

$200 Million

Snowflake is planning $200 million in an AI Hub and startup investments.

Reply

or to participate.