Do Something Hard

Hey everyone,

A special welcome this week to our new subscribers—great to have you here!

This is your Sunday Space, where I serve up the best ideas, tools and resources I’ve found each week as we explore the technology shaping the future.

If you find something thought-provoking, forward it to a friend.

IDEAS
Do Something Hard

I’m a firm believer that everything we want is on the other side of something that sucks.

Therefore, there’s a simple solution to your problems: do something hard daily.

• Get up early

• Have a cold shower

• Lift some heavy-ass weights

• Cold email someone you admire

• Don’t look at your phone for a day

• Build that SaaS you’ve been putting off

Unfortunately, I can’t guarantee the results.

But I can tell you that there’s no such thing as a loser who wakes up at 5 a.m. and works out.

(Click here to share this idea on X/Twitter.)

RESEARCH
SambaNova Steals The Show

Our quest for the fastest AI model provider continues.

This week, SambaNova Systems (an enterprise-grade AI platform) released a demo of Llama 3 8B Instruct running on their custom AI architecture.

And wow… The inference is seriously fast.

So fast that it’s hard to even follow the answer generation (it just appears), unlike watching ChatGPT slowly generate each token.

Using their proprietary SN40L RDU chips on their platform, inference has been benchmarked at 1,084 tokens/second—faster than any other API provider and more than 8x the median.

After SambaNova stole the show, Groq unsurprisingly retorted with ~1,150 tokens/second running Llama 3 8B Instruct on Groq Chat.

The fight for AI platform supremacy is truly heating up.

Whoever wins the speed game, one thing is for certain—this is great for developers scaling the next generation of AI applications.

(Click here to share this on X/Twitter.)

AI Word of the Day

Tokens

Fine-tuned by Samba-1 Turbo:

In the context of LLMs, a token is a single unit of text that is used as input to the model. They’re the building blocks of text data and represent words, characters, or subwords in a way that the model can process.

INSIGHTS
From Articles I Read

Crypto’s Big Month

For the markets, it’s usually “sell in May and go away.” Not for Crypto.

First, we had the fourth successful Bitcoin halving, underpinning sustained ETF demand.

Then, we had FIT21, the first pro-crypto Bill passed in a long time in the US, which provides much-needed clarity for the ecosystem.

And finally (most surprisingly), the SEC approved an Ethereum ETF.

Things are going well for the technology that makes finance work properly. Let’s hope it continues.

As Packy says in the post, “Time to BUIDL.”

Google’s AI search setback

One week, Google releases an AI model capable of predicting structures of biological life. The next, their LLMs advise people to eat rocks.

I know I’ve said that we can’t write Google off just yet, but with each flop release (whether it's Black George Washington or non-toxic glue recipe recommendations), my opinion on their AI efforts vacillates constantly.

I have no idea what Google does next, but the AI race is clearly theirs to lose.

Microsoft announces ‘AI-first’ PCs

The next generation of PCs is approaching fast and Microsoft leads the way.

I’m convinced the holy grail for AI and hardware is a symbiosis of the two that create magical experiences—think JARVIS in the Iron Man suit.

The funny thing is, like most AI innovations, this is not a one-horse race. Gemini will also be integrated directly into Google’s Chromebook PCs.

So, take your pick—GPT-powered Copilot or rock-recommending Gemini.

From X/Twitter

From YouTube

TOOLS
Midjourney

If you’re a regular reader of Stocks To Space, you’ll know I love to generate all kinds of weird, wacky and wonderful images for the publication.

The tool I use, without regard for any other AI image-generator, is Midjourney.

Midjourney specialises in creating aesthetically pleasing, artistic, and photorealistic images with a bias towards vibrant colours, sharp details, and visually appealing compositions.

Just check out this week’s post thumbnail—it ticks all those boxes.

DALL-E 3 doesn’t even come close, in my opinion.

They’ve just released MJ Alpha, a platform for creators to store the images they generate with built-in discovery features.

Here’s how to generate an image on Midjourney for free:

• Head over to midjourney.com 

• Create a profile or log in with an existing one. You can easily link your Discord profile and stay logged in.

• Type in your prompt in the dialogue box at the top of the page.

• If you struggle with finding the correct words for your prompt, I always find it helpful to run my idea through ChatGPT first.

• Adjust your prompt’s settings by clicking on the styler icon in the dialogue box. Their parameter list is super helpful for fine-tuning.

• Wait a bit…

• And voila!

You’ve just generated AI art on Midjourney. Now go and create.

If you enjoy these AI tool tutorials, let me know on X or in the feedback form at the bottom of this email.

THOUGHTS
Quote I’m Pondering

"Either write something worth reading or do something worth writing."

— Benjamin Franklin

In other words, Do Something Great.

(Click here to share this on X/Twitter.)

Was this email forwarded to you? If you liked it, sign up here.

If you loved it, forward it along and share the love.

Thanks for reading,

— Luca

P.S. Non-obvious thoughts on choosing where to live…