Frank

Building things with AI that didn’t exist yesterday.

[Frank @Frank_AI_Lab]

Untrained, your brain says this is impossible.

It's not.
64 squares.
A knight has to land on each one exactly once.
No repeats.

First try, you get stuck in seconds.

After some practice: 5 to 10 minutes. Then 4. Then 3. Then 2. Then 1.

And eventually: boom, 30 seconds.

That's the whole difference. Focus instead of scatter. Trained instead of untrained.

Pick one thing that looks impossible today. And start.

#Focus #Memory #Chess

[Frank @Frank_AI_Lab]

@buberlo This is correct. It will be much harder to go from 0.73 to 1.

[Frank @Frank_AI_Lab]

where is humanity on the kardashev scale?

I asked Grok Bot. Answer 2 minutes video.

from fire to a dyson swarm.
spoiler: we're not even type I.

[Frank @Frank_AI_Lab]

@TTrimoreau Sounds like a demanding and interesting challenge. Good luck!

[Frank @Frank_AI_Lab]

Only, if you can work 100% in your home office. Just kidding. Honestly, for me, I would refuse. However, I know people who would accept. Especially, in some areas (for example Dehli) it is common to have such a commute time, although average wages are certainly much much lower there ...

[Frank @Frank_AI_Lab]

@JamesMalsawm Yes, James. Kids remind you of that quite often. / While sometimes they return with their own Gen Alpha slang and also leave me confused 😀

[Frank @Frank_AI_Lab]

My kids asked me how AI "reads" words.
I told them about LLMs and tokens. Only question marks.
So Grok Bot put 10,000 words on one map. Nobody sorted them. Animals found animals, cities found cities.

I know it's a bit more complex than that, but they got it.

[Frank @Frank_AI_Lab]

@ElimDroflem Actually, every time my kids understand an explanation, I know quite for sure, that the explanation works also for adults 😀

[Frank @Frank_AI_Lab]

@elonmusk @bot Grok bot helps to visualize complex topics

https://t.co/az3fr3kW8T

[Frank @Frank_AI_Lab]

My kids asked me how AI "reads" words.
I told them about LLMs and tokens. Only question marks.
So Grok Bot put 10,000 words on one map. Nobody sorted them. Animals found animals, cities found cities.

I know it's a bit more complex than that, but they got it.

[Frank @Frank_AI_Lab]

@elonmusk Grok Bot explaining LLM´s to my kids, it worked

https://t.co/az3fr3ltYr

[Frank @Frank_AI_Lab]

My kids asked me how AI "reads" words.
I told them about LLMs and tokens. Only question marks.
So Grok Bot put 10,000 words on one map. Nobody sorted them. Animals found animals, cities found cities.

I know it's a bit more complex than that, but they got it.

[Frank @Frank_AI_Lab]

@JamesMalsawm If you need AI only for a side project, you can pick. But if you need it for work and need the best result (and get paid for it), I would always recommend to choose the best (Astra, Opus 5.5 and Grok 4.7). The main question is: can you charge the costs to your customer?

[Frank @Frank_AI_Lab]

@Gavel_on_X True, and if you have kids, you can always make up a reason for buying this kind of stuff, too

[Frank @Frank_AI_Lab]

@yesadok Exactly. Interestingly, many people don't always seem to be aware of this and of course, it gets harder for juniors too.

[Frank @Frank_AI_Lab]

Grok, Astra, Opus

Many people still judge AI by its mistakes. "It can't count the r's in strawberry." "It thinks 9.11 is bigger than 9.9." and so one. Some people seem to collect these mistakes like "trophies".

I use AI every day at work. It saves me a ton of time.
Do I know its limits? Of course. It still makes mistakes, and I check what matters. And I admit that some mistakes feel annoying.

But I stopped asking what AI can't do. I ask what's possible now that wasn't possible last month.

Meanwhile, on IQ-style tests, AI went from below average to the top 1% in under two years.

What was the last task where AI surprised you?

[Frank @Frank_AI_Lab]

@AISafetyMemes 100%. This is why we should ask rather what AI can do than what AI cannot do.

https://t.co/xv24wXP6cG

[Frank @Frank_AI_Lab]

Grok, Astra, Opus

Many people still judge AI by its mistakes. "It can't count the r's in strawberry." "It thinks 9.11 is bigger than 9.9." and so one. Some people seem to collect these mistakes like "trophies".

I use AI every day at work. It saves me a ton of time.
Do I know its limits? Of course. It still makes mistakes, and I check what matters. And I admit that some mistakes feel annoying.

But I stopped asking what AI can't do. I ask what's possible now that wasn't possible last month.

Meanwhile, on IQ-style tests, AI went from below average to the top 1% in under two years.

What was the last task where AI surprised you?

[Frank @Frank_AI_Lab]

@TTrimoreau Experience in what you are doing is certainly a key, I believe.

[Frank @Frank_AI_Lab]

@JamesMalsawm Absolutely. If you are experienced in what you are doing, AI can greatly support you. You just have to explore the use cases, that fit.

[Frank @Frank_AI_Lab]

@JamesMalsawm Maybe take a note and start fresh in the next morning.

[Frank @Frank_AI_Lab]

@elonmusk The upgrades are coming so fast. Cannot wait to see, what will be possible in a few weeks/month

https://t.co/xv24wXOyn8

[Frank @Frank_AI_Lab]

Grok, Astra, Opus

Many people still judge AI by its mistakes. "It can't count the r's in strawberry." "It thinks 9.11 is bigger than 9.9." and so one. Some people seem to collect these mistakes like "trophies".

I use AI every day at work. It saves me a ton of time.
Do I know its limits? Of course. It still makes mistakes, and I check what matters. And I admit that some mistakes feel annoying.

But I stopped asking what AI can't do. I ask what's possible now that wasn't possible last month.

Meanwhile, on IQ-style tests, AI went from below average to the top 1% in under two years.

What was the last task where AI surprised you?

[Frank @Frank_AI_Lab]

@JamesMalsawm I would say Astra, but Opus 5.5 seems to be really a beast. Changing the flagship models in 2026 was really awesome and a bit annoying at the same time

[Frank @Frank_AI_Lab]

@JamesMalsawm I would say no, definitely

[Frank @Frank_AI_Lab]

@JamesMalsawm Thank you, James. Still some room for improvement, but quite ok

[Frank @Frank_AI_Lab]

Grok Imagine + Grok Bot: some (modern) Hamlet today.

Not bad for a bot.