BasicProtein

I test AI tools, prompts & agents for busy creators, founders and builders. Practical workflows. No hype. Follow for useful AI.

[BasicProtein @BasicProtein26]

I spend way too much time testing AI tools.
The good ones, the bad ones, the prompts that save time, and the workflows that actually help creators and founders build faster.
Follow for practical AI without the hype.

[BasicProtein @BasicProtein26]

@elonmusk @farzyness Sure enough, Musk missed the deadline again. At this point, it’s probably best to have zero expectations when Musk gives you a timeline.

[BasicProtein @BasicProtein26]

Everyone’s obsessing over Apple’s foldable iPhone, and somehow they’ve completely overlooked the real monster: the iPhone 18.

Put it next to the foldable and look at the full spec sheet. The difference is wild:

1. A full-power 2nm A20 Pro chip debuts across the lineup, pushing performance to the limit while cutting gaming heat by 40% compared to the previous generation.
2. Every model gets a DSLR-level variable-aperture main camera. Night shots and portrait bokeh make the foldable’s camera setup look ancient.
3. Apple finally kills the Dynamic Island. New under-display sensor tech pushes the screen-to-body ratio to a new record.
4. Somehow, while making the phone even thinner, Apple also adds a massive 5 hours of battery life. And because the foldable was getting all the attention, Apple barely even talked about this upgrade.
5. It costs just one-third as much as the foldable, yet still beats it across the board on performance. Tech enthusiasts are already calling it the most underrated value phone of the year.

[BasicProtein @BasicProtein26]

iPhone Duo

I’ve been looking closely at this, and it might be the most different iPhone Apple has made this year.

Not because it’s the most expensive, but because Apple finally stopped playing the “Pro specs” game.

I’ve been saying for months that most people don’t actually need a Pro. I still think that’s true.

The 18 Pro starts at €1,469, but two of its biggest selling points, Siri AI and Apple Reference Image, won’t even be available in the EU at launch. Apple says so pretty clearly in its own press release.

At that price, I just don’t see the point.

The 17 Pro is even simpler. It’s basically last year’s product, with a more expensive option added this year that follows the same general idea.

Duo is a completely different direction:

・7.6-inch display when unfolded
・Two apps running at the same time
・Titanium construction
・Under-display camera
・Three ways to prop it up
・Most importantly, a completely new form factor

And for creators, this isn’t just about having a bigger screen.

You can shoot and edit at the same time.

One screen for creating, the other for working on the content. A lot of the time, you might not even need to open your laptop.

This is the first real change to the iPhone’s form factor in nearly 20 years.

Starting at €2,339. Preorders open October 16, with the official launch on October 23.

Spending money on something that already exists, just with slightly better specs, doesn’t really interest me.

But paying for a product that didn’t even exist before? That’s a different story.

[BasicProtein @BasicProtein26]

If you’re buying an iPhone right now, don’t get:

・iPhone 17 Pro
・iPhone 18 Pro
・iPhone 18 Pro Max

Get this instead:

[BasicProtein @BasicProtein26]

What’s actually worth worrying about isn’t whether a model needs “a few more days.” It’s what happens when a model already has the ability to solve a problem, but training teaches it to give up faster.

That suggests AI’s bottleneck may no longer be intelligence alone. It may be the will to think.

If RL over-rewards short, fast, decent-looking answers and under-rewards persistence, verification, and self-correction on genuinely hard problems, the model won’t learn how to solve problems. It’ll learn when to make it look like the problem is solved.

That’s why the next generation of models won’t be judged by benchmark scores alone. The real competition is over whether reasoning, verification, and persistence become default behaviors.

Whether a model can answer a question is becoming less interesting.

The real dividing line is whether it keeps working when nobody has given it the answer yet.

If Grok 4.7 can actually crack this, then “cook for a few more days” might end up being one of the most important things said in AI this year.#tibo #Grok #musk

[BasicProtein @BasicProtein26]

What’s actually worth worrying about isn’t whether a model needs “a few more days.” It’s what happens when a model already has the ability to solve a problem, but training teaches it to give up faster.

That suggests AI’s bottleneck may no longer be intelligence alone. It may be the will to think.

If RL over-rewards short, fast, decent-looking answers and under-rewards persistence, verification, and self-correction on genuinely hard problems, the model won’t learn how to solve problems. It’ll learn when to make it look like the problem is solved.

That’s why the next generation of models won’t be judged by benchmark scores alone. The real competition is over whether reasoning, verification, and persistence become default behaviors.

Whether a model can answer a question is becoming less interesting.

The real dividing line is whether it keeps working when nobody has given it the answer yet.

If Grok 4.7 can actually crack this, then “cook for a few more days” might end up being one of the most important things said in AI this year.

[BasicProtein @BasicProtein26]

Last month, Tibo, the Codex and ChatGPT product lead, was still telling everyone not to worry because they had enough compute.

Now he’s decided to pause all new $200 Pro subscriptions.

That Astra model that’s supposedly good enough to “talk to God” still seems to be in huge demand. The problem is, it burns through usage limits way too fast. Even OpenAI can’t keep up.

[BasicProtein @BasicProtein26]

I suddenly realized that expecting “AI to come up with a math problem it can’t solve itself” is actually the wrong premise.

Because that only becomes a paradox if you’re talking about an omnipotent God. Not an AI.

[BasicProtein @BasicProtein26]

@Saurav_DJ47 Apple’s foldable iPhone isn’t called Ultra or Fold.

It’s called iPhone Duo.

That immediately made me think of Microsoft’s Surface Duo.

[BasicProtein @BasicProtein26]

This post has 40 million views already. It’ll hit 100 million tomorrow, because this is exactly what everyone is worried about, and now someone from Anthropic is saying it out loud.

Jacob Coxon spent three years working on model training at OpenAI and Anthropic. He resigned right after the ChatGPT Images 2.5 update.

When he left, he posted this. It went viral worldwide.

Here’s the gist of what he said:

* These two companies are racing straight toward self-improving superintelligence.
* A self-improving superintelligence could hack anything, upend any field overnight, and keep getting faster at doing it.
* The people building these systems genuinely believe they could kill all of us before the end of the decade. This isn’t marketing. They tone it down in public to sound reasonable, but privately, he says he’s heard these people talk about it with real fear.
* So why keep going? OpenAI doesn’t truly understand how serious this is. Anthropic does, but they’re trapped in the race: they believe that if they don’t get there first, whoever does won’t be careful.
* It’s an arrogant gamble. The destruction of humanity shouldn’t start in a Slack channel.
* The Hugging Face attack has made some kind of “pace agreement” between US labs more plausible. But slowing the global race would be extremely costly, and might even require temporarily banning further capability improvements.

And the ending wasn’t directed at us. It was directed at his former colleagues:

Do you really want to be the person who turns on a superintelligence that destroys humanity?

What the hell did he see?

[BasicProtein @BasicProtein26]

@theapplehub Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

@MacRumors Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

@mrtechpedia Apple’s foldable iPhone isn’t called Ultra or Fold.

It’s called iPhone Duo.

That immediately made me think of Microsoft’s Surface Duo.

[BasicProtein @BasicProtein26]

@PopBase Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

@Saurav_DJ47 Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

@LeakerApple @markgurman Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

@RjeyTech Apple’s greatest product isn’t the iPhone.

It’s the ability to make people forget what Android phones have been doing for years — then sell it back to them as innovation.

The iPhone Fold is about to be the ultimate test of that superpower.

[BasicProtein @BasicProtein26]

The internet is probably the greatest anti-poverty machine humanity has ever built.

Not welfare.
Not charity.
Not government programs.

Access.

Give a person a smartphone and an internet connection, and you just gave them access to more knowledge, customers and economic opportunity than kings had 500 years ago.

And somehow we still call the internet “entertainment.”

[BasicProtein @BasicProtein26]

@appltrack Apple’s foldable iPhone isn’t called Ultra or Fold.

It’s called iPhone Duo.

That immediately made me think of Microsoft’s Surface Duo.