#126: How to Use Jev, Generating 175M Impressions, Learn Cloud Engineering Fast

Hi All!

We are pleased to welcome you to this week's edition of Undiscovered, a newsletter with exclusive resources and insights expanding from the material found on our main site - becketu.com.

This edition is brought to you by The Morning Brew. Be sure to check out their exclusive offer in the P.S.

This week we will take a look at how to best use Jev AI, the strategy behind a company's 175M impressions within the last year, teaching yourself cloud engineering, and more. Let's dive in:

What is Jev & How Can You Use It?

​Matt Canham wrote the clearest guide yet for understanding the newest viral tool, Jev. The model is built specifically to make decisions rather than generate text, and people are excited about its cost-effectiveness. Because it returns a single answer with a confidence score instead of writing out a full response token by token, it runs 20 to 200x faster and 40 to 400x cheaper than a comparable LLM.

Using it comes down to giving it three things:

1. The question you want answered
2. The set of possible answers
3. Context it needs to decide (which it calls "state")

It's primarily used to handle three types of decisions:

1. A yes or no question, like whether a customer message is a complaint
2. A choice where it picks from up to 255 options, like which team should handle a request
3. A score where it rates something on a scale, like how satisfied a customer is.

The reason this matters is that every product already runs on hundreds of small decisions, like whether a payment looks fraudulent, whether a user is about to cancel, or which tutorial to show next, and right now those are either hardcoded with rigid rules or skipped entirely because running a full LLM on each one costs too much.

Jev is priced so you can put a decider on all of them at once, and a separate builder designing agents around the same principle points out why. The expensive part of AI has been re-reading all the context to produce a written answer, so a model that just decides is cheaper.

Visualizing Disk Space & Creating Dream Software

​Tobi Lütke, the CEO of Shopify, was wondering where his disk space went, so instead of hunting for an app he just made one that beautifully shows his storage and posted the tool free on GitHub. "You can simply wish software into existence."

​Jason Fried, the founder of 37signals, had wanted a specific writing tool for a decade, something that let him play with alternatives at the word, sentence, and paragraph level, and it never made sense to build because it was not commercially viable and would have taken months the old way. This weekend he just made it himself with Claude's help. He calls it Write_On, and the detail that matters is the reason it exists. It was too niche to justify a company's time, so for ten years it stayed an idea, and now it took a weekend.

Insights After 175 Million Impressions in a Year

​The Threadguy Stream broke down how they became one of the largest shows in finance with over 175 million impressions in their first year, and the core insight is that their livestream is the product but their real business is clips. A single three-hour stream reaches about 7,000 live viewers, but it then generates 850,000 downstream impressions because they chop it into 100+ pieces of content across 150+ accounts on every platform.

They treat distribution as a separate discipline from creation, with a tiny team deciding what the show believes and a huge network of clippers figuring out how to spread it. Their phrase is "centralize taste, decentralize distribution". They also think in terms of shots on goal.

They cannot predict which clip will go viral, so instead of trying to be right once, they make the best stream they can and then give it a hundred chances to hit by packaging it every possible way. Content quality gives them an edge, and clipping gives that edge as many opportunities as possible to express itself.

Teach Yourself Cloud Engineering

​Polymarket Money reported that Anthropic is reportedly planning to spend $518,000,000,000.00 on cloud and infrastructure in the coming years.

​GPS tweeted out a resource they made for anyone looking to take advantage of the opportunity - learntocloud.guide. It's a free, open-source program for teaching yourself cloud engineering.

It works like a structured course but it is self-paced. You follow a curriculum with resources and practical challenges, submit your work for automated checks and AI-assisted reviews, and build real things from Linux and networking up to deploying and securing an AI-powered API on a cloud of your choice. People working in the field keep the material current so you are dynamically learning how cloud engineering is actually practiced today.

Agents Are Eating All the Apps

​Ben Thompson's latest Stratechery piece argues that agents are the ultimate aggregators, and his clearest point is that apps 'were never the point'. You do not have a million apps on your phone because you love apps. You have them because each one was the only way to get a specific job done, and you had to learn how each one worked to do it.

Agents collapse the old system. Thompson made a new app for himself just by asking his agent to organize the recipes he had saved on Instagram, and it took five minutes while he walked his dog. Once an agent can operate any app or website for you, our old ways of interfacing with apps becomes obsolete.

Models are substitutable, but agents get stickier the more they know about you, your logins, your files, your habits, so most people will end up with fewer agents. That makes the fight to be the agent you talk to the biggest prize in tech, and it is why the companies with distribution, Meta reaching every person and Microsoft reaching every employee, are the ones positioned to hand you an agent with a computer already attached.

We appreciate any feedback you are able to provide here. What do you want more or less of? Other suggestions? Feel free to reach out to us on Instagram and give us a follow there, tag your friends on our posts, and please forward this newsletter along to anyone else who would enjoy it.

Disclaimer: Becket U is an Amazon Associate and purchases through Amazon links may earn a small affiliate commission, but the price is the same for you. We only recommend books we love and think you would love, too.

Always wishing you the best,

J.B.