Big Idea

The argument in one line.

Decision models like Jev trade conversation for speed: by only returning a yes/no, a choice or a score, they let code make fast, cheap AI judgments that chat models are too slow and costly to do.

Who This Is For

Read if. Skip if.

READ IF YOU ARE…

A developer or vibe-coder wiring AI into an app who is paying for slow, expensive chat-model calls just to classify, route or filter text.
Someone who keeps seeing Jev, Clef and the OpenAI Decisions API on X and wants a plain explanation of what a decision model is before deciding whether to adopt one.
A builder running coding agents who wants concrete ideas like model routers, file search and live script tracking that a fast decision primitive makes affordable.

SKIP IF…

You only chat with AI assistants and never write or generate code that calls a model.
You need open weights or a known training set, since Jev is closed and its data is undisclosed.

TL;DR

The full version, fast.

Jev from TypeSafe is a decision model, a 'system one' AI that cannot talk or call tools and only returns yes/no with confidence, a multiple-choice pick, or a score. Because it skips natural language, it classified a spam comment in 0.17 seconds versus about two seconds for ChatGPT, and handled 100 comments far faster and cheaper. It acts like an AI-powered if statement that needs no retraining, unlike a classic classifier. NetworkChuck shows it routing models, driving a live teleprompter and ordering pizza, and in a 107-comment benchmark it edged OpenAI's Decisions API and beat local models on accuracy and cost.

Free for members

Chat with this breakdown — free.

Sign in and you get 23 free chat messages on us — ask for the hook, quote a framework, find the exact transcript moment, generate a markdown action plan. Bring your own key when you want unlimited.

Create a free account →

Chapters

Where the time goes.

00:00 – 00:51

01 · What is Jev?

Cold open: a model that cannot talk, with copycats from OpenAI, Cloudflare and Ollama.

00:51 – 02:00

02 · Jev can't talk

Playground demo: a spam comment is judged 92% true and the output is JSON, not words.

02:00 – 04:09

03 · Sponsor: Twingate

Sponsored read worked into a Jev question about safe remote access for AI agents.

04:09 – 05:21

04 · Jev vs ChatGPT speed

0.17 seconds versus about two seconds, and the three answer types: yes/no, choice, score.

05:21 – 05:40

05 · 100 comments test

Batch run shows the gap in speed and cost at volume.

05:40 – 08:51

06 · Why Jev is different

Jev talks to code, not people. TypeSafe cofounder explains system one models.

08:51 – 11:16

07 · Jev vs classifiers

Smart if statements, the password example, and why no retraining is the edge over classifiers.

11:16 – 12:09

08 · How do we use it?

Who it is for: developers, which now includes anyone generating code with AI.

12:09 – 15:08

09 · Model router and Intelliprompter

Router demo picking small, medium or large models, then a live teleprompter that checks off script bullets.

15:08 – 15:53

10 · Ordering Domino's with Jev

Browser agent orders pizza in 15 seconds versus about 61 seconds for Codex.

15:53 – 16:17

11 · JevGrep, Hermes and Doom

Community projects: file search cutting agent cost, a Hermes plugin, and Jev playing Doom.

16:17 – 17:16

12 · Jev vs other decision models

Benchmark on 107 fan comments against OpenAI Decisions API, Clef and local models.

17:16 – 19:03

13 · Final thoughts

Docker and Lambda analogy for why decision models may become a standard primitive.

19:03 – 20:35

14 · Prayer

Host's closing prayer, unrelated to the technical content.

Atomic Insights

Lines worth screenshotting.

A decision model returns a structured judgment instead of prose, so it is built to talk to your code, not to you.
Jev scored a YouTube comment as 91% spam in 0.17 seconds, while a chat model took roughly two seconds on the same task.
System one models give the instant gut call, system two models reason step by step, and most LLMs are system two.
Using a reasoning chat model to check whether a password is strong is like using a chainsaw to slice bread.
A classic classifier only answers the one question it was trained on, while a decision model accepts a new question with no retraining.
Critics call decision models 'AI powered if statements', and the builders agree: smart switch statements and logic gates that choose which path to take.
Jev only has three answer types: noul (yes/no with confidence), choice (multiple choice) and score.
Because coding agents mean everyone writes code now, a primitive you do not understand is fine as long as your agent can wire it in.
A model router that sends '2 plus 2' to a small model and a proof of infinite primes to a large one saves money without a slow reasoning step.
A live teleprompter that checks off script bullet points every half second went from about $30/hour to reasonable once a decision model did the checking.
In a 107-comment test, Jev was most accurate and cheapest in the cloud, OpenAI's Decisions API was close behind at 2.4 times the cost.
The best local result came from the 27B Clef model, but cloud decision models still led on accuracy.
Decision models may be at the Docker and AWS Lambda moment, hard to grasp until everyone sees what they enable.

Takeaway

Fast decisions beat fluent answers for automation

WHAT TO LEARN

When code, not a person, consumes the AI output, a model that returns only a yes/no, a choice or a score is faster and cheaper than a chat model.

02Jev can't talk

A decision model gives you a verdict as JSON with a confidence value, not an essay, so the output can feed straight into code.

04Jev vs ChatGPT speed

Speed and cost gaps grow with volume, so batch jobs like moderating 100 comments are where decision models pay off.
The three answer types, yes/no, choice and score, cover most gating, routing and rating tasks.

06Why Jev is different

Chat models were tuned to be assistants. Decision models are tuned for background automation and boring work.
System one models fit instant calls, while system two reasoning models are overkill for simple checks.

07Jev vs classifiers

A decision model matches a classifier's speed and near its accuracy but accepts new questions without retraining.

09Model router and Intelliprompter

Routing easy prompts to small models and hard ones to large models cuts cost without adding a slow reasoning step.
Checking whether an idea was discussed, not exact words, made a live teleprompter affordable.

12Jev vs other decision models

On 107 real comments Jev was most accurate and cheapest in the cloud, and OpenAI's Decisions API cost 2.4 times more.

Glossary

Terms worth knowing.

Decision model: An AI model built to return a structured judgment, such as yes/no with confidence, a choice or a score, instead of conversational text.
System one model: A model optimized for instant gut calls, borrowing Daniel Kahneman's term for fast thinking, as opposed to slow step-by-step reasoning.
Classifier: A machine learning model trained on a labeled dataset to answer one fixed question, like whether a password has leaked.
Model router: A layer that reads each request and sends it to a small, medium or large model depending on how hard it looks.
Closed weights: A model whose trained parameters are not published, so you can only use it through the provider's API.
noul: Jev's yes/no answer type, returning true or false along with a confidence value between 0 and 1.

Resources

Things they pointed at.

02:00productTwingate ↗

01:00productJev (TypeSafe)

00:00productOpenAI Decisions API

00:00productClef (Cloudflare)

16:00toolJevGrep

00:00linkJev guide repo ↗

08:00bookThinking, Fast and Slow

Quotables

Lines you could clip.

00:20

“And that's how you know something is good.”

short, punchy social proof line→ TikTok hook↗ Tweet quote

08:30

“It'd be like using a chainsaw to slice a piece of bread.”

vivid analogy, no setup needed→ IG reel cold open↗ Tweet quote

09:30

“It's like a logic gate that uses AI to decide which path to go.”

one-line definition of a decision model→ newsletter pull-quote↗ Tweet quote

The Script

Word for word.

Read-along

Don't just watch it. Burn it in.

See every word as it's spoken — crank it to 2× and still catch all of it. The same dual-channel trick behind Amazon's Kindle + Audible.

Jev is a new AI that seems kind of dumb at first. It can't call tools, it can't search the web, it can't even talk. But everyone is freaking out about it.

Shoot, everyone is copying it. And that's how you know something is good. Decisions API from OpenAI, Clef from Cloudflare, even Ollama.

And to add to your confusion, here it is right now playing Doom insanely fast. Also ordering a Domino's pizza in seconds. What?

is this let's figure it out how can you automate anything if you get like Einstein 95 % of the time and mr. being the other 5 % in this video we'll talk about what Jeff is and what a decision model is and how can you use it or even if you should is this for you also I am literally using Jeff right now like as we speak or as I speak it's amazing you'll see so get your coffee ready it's time to learn about some system one models So Jeff can't talk.

And that's one of its biggest selling points. What you see when you talk to an LLM, like chat GPT or Claude, you input words and it will output words. We're used to this, but when you talk to Jeff, you do input words, but the output is not words.

It's a decision, which I know it sounds weird. You just have to try it. Let's try it together.

We'll open the Jeff playground, which is now open to everyone for a while. They had it closed because it was so popular. And we'll ask it a simple question.

Is this YouTube comment spam? And we'll put that right in here in the question box. But what YouTube comment, this one right here, this is a real one straight from my comment section.

By the way, if you do wanna use this example, I have it all in the description below and notice we're putting this in the state box. Now let's run it. Here's the run button down here.

Ready? Set Jev boom. Now I didn't speed that up.

It was that fast. And if I scooch this over, you'll see, this is how long it took. And it was like, yeah, it's about 92 % true that that comment is spam.

Now at the top here, if we switch over to Jason, this is exactly what Jeff is delivering to us. It's not talking to us. It's just giving us a decision.

But what does that mean? What's the difference? Why should we care?

And before we do that, I have one very important question to ask Jeff. And this question's actually on behalf of you watch. The state is I want my AI agents to reach a few things from my network, my home lab from anywhere, but I don't want to give them access to everything.

And I don't want to open any ports. This is you, right? You got some cool stuff running in your lab and you wanna give stuff access to it.

And you're tempted just to open a port. And here's the question is twin gate a good fit for this person. Yeah.

You guessed it. Twin gate is a sponsor of this video. That was my attempt to integrate this ad and not make you wanna skip it.

Don't do it. Please. How did I do?

I mean, for real, this is probably gonna help you because you are probably this person trying to open up ports. You're doing some cool stuff. I get it.

Don't expose it. Let me give Jeff some context. Here's what twin gate is secure remote access with no open ports.

Give a person a device or AI agents access to only the exact things you choose. You can't do that with VPN, by the way.

This is free for home labs. I hope it understands the extra ease. We'll see what Jeff says.

Should you use it? Dude, you gotta listen to Jeff 85 % true. That means you have to use it.

It's a decision model. And it made that decision for you. Now this AI agent thing is a true story here.

I've been playing with grock bot and this is a cloud only agent. It lives up somewhere in Elon Musk's house. And I wanted to give grock bot access to some stuff I have inside my network.

Not everything. I'm not crazy, but I have some custom apps that I don't expose to the internet, but I wanted grock bot to use it. And twin gate gives the most amazing way to do this.

They have this thing called a service account. And when I set it up, I can say you can only access this one custom app and nothing else in my network. And when I give this to Grokbot, that's all it can access.

But it's so sick because my cloud -based AI agent can access one thing inside my home network, and that's stinking it. And to set it up is stupid easy. All you need is one thing running inside your network.

It can be a Raspberry Pi, it can be something on your NAS. A TwinGate connector can run on anything, pretty much. Yeah, pretty much anything.

And once you have that running twin gate takes care of the rest. This is like an enterprise level way to secure yourself. And it's free for home labs.

Jev said you have to use it. The decision's already been made. Check the link in the description.

Well, let's paste the same thing inside chat GPT. I put this into a custom interface, by the way, ready, set, go.

Two seconds said it was spam, but over here in Jev again, go 17 seconds, 91 % sure. The speed difference is kind of crazy and I'm using GPT six Luna instant. Why is it faster?

Because Jeff is making a decision instead of giving us a long drawn out answer. He literally cannot talk to you the same way that chat GPT and claw do. He's actually limited to these three response types, new, which we just saw essentially.

yes or no, and how confident it is on that. Yes or no. We could actually ask more than one question at once.

So look at all these newels I have asking ready, set, go. See how fast that was. Another one is choice, multiple choice, asking the question with these options.

Let's try it out.

And pretty much instantly it told us that it was trying to promote something 98 % sure. And the third one is score. How positive is the comment negative, neutral or positive?

Let's run it.

And I rated the comment as positive. So what? Jeb is giving us a simple answer.

Can't chadjbt just give us a yes or no answer back? Wouldn't that be faster? Let's try it back to my custom interface.

Let's try chadjbt.

Oh, yikes. That was slow. Yes.

And then Jeb again. So I think you see it. And this really becomes powerful when you think about not just classifying one comment, but what if we had to do a lot?

Like what if we did 100 real comments for my channel? Let's try it out. 100 comments.

Ready? Set. Go.

Come on. Still waiting.

So right off the bat, we noticed two things, right? Speed and cost. ChadGBT, Claude, and all these other LLMs like it were built to respond to us with natural language.

That's what they were optimized for. But Jev, it was built different. They were autocompletes.

And ChadGBT made them incredible at being an assistant. And like we're kind of like doing an orthogonal jump where instead of making it incredible at being an assistant, we're making incredible at like run in the background automation, like do like the boring work type tasks. Jeff isn't talking to you.

It's talking to code. People would sit there and they would think of it like a chatbot, like a chat GPT, and they would go. okay, so now that this is in my life, what questions do I have for Jev?

And it's like, no, it won't be used that way. You're not going to ask it, is this a good idea for me to do? You're not going to go to it and ask it like it's an oracle.

It is going to be used inside code. And that's what Jev is optimized for. Well, I think TypeSafe is here to correct, I will say.

And I don't blame any single person except for the AI labs. I think that... over -promise, under -deliver has been a massive issue.

It is not anyone who believes them because, like, what better do you know? But, like, there's many shapes of AI. AI is, like, just an interesting, super awesome technology and there's so many more ways to expose it to the world than simply, like, building a machine god in a data center and having it be, like, the next species on Earth.

That is the real closing that's happening because we actually want AI to be... reliable. And by the way, this is Diogo Almeida co -founder of type safe.

He actually worked at open AI on instruct GPT, the research that led to chat GPT. And he left open AI because he thought AI could do more. And by the way, this was the more he was talking about.

And the funny thing is we don't really know what makes Jev better at this. It's closed weights and we don't know what data it was trained on. And that is kind of the secret sauce of Jeff.

All that we do know is that it was optimized to do this one thing, make decisions. They call this a system one model. It's an idea from Daniel Kahneman's book.

It's called thinking fast and slow system. Two is slow. Step -by -step reasoning system.

One is that instant gut call like, yes, don't think about it. Just do it. And I'm sure you guess by now that most LMS are kind of meant and built to be system two type things.

They think longer, they reason, and that's good for a lot of stuff. But the thing is you really don't need system two for basic. quick decisions like is password one, two, three, a good password.

It'd be like using a chainsaw to slice a piece of bread. We don't need something that heavy, but like all new things. Um, not everyone's a fan.

Not everyone was like, oh yeah, Jeff is awesome. Many claim that it's just like an AI powered if statement. And you know what?

They're not exactly wrong. a way to think about typesafe especially when trying to explain it to somebody brand new is to think of it as like smart if statements or smart switch statements right like ai infused logic gates in fact it was my kid my 11 year old that i was trying to explain what our like model is to them a week before we launched and when i like explained it in a much longer thing they're like oh so it's like a logic gate like it's like a logic gate that uses ai to decide which path to go let's take our password example with password123 you might use code to determine if that was a strong enough password for your user that could be an if statement it might look like this if it's over eight characters and has an uppercase and has a number and has a symbol it's a strong password i think you and i both know that's not a strong password so in this case we might use something else called a classifier in our example it could be a special centralized machine learning model trained on leak passwords.

And this would actually be really great at detecting it as a bad password. And as you can see here, Jeff can do the same thing, not a strong password. So yeah, Jeff and the classifier are doing the same job.

So how are they different? Why did we need Jeff? We already had classifiers, these machine learning things.

Well, the classifier only knows one question. We would have to train it on new data to ask a different question with Jeff. It doesn't have to be trained again.

You can simply just ask it another question. It'll be just as fast and nearly as accurate as a classifier without needing a machine learning engineer to set up the data set for you and everything. I love calling Java classifier.

People think it's an insult, but only machine learning people think it's an insult. So. Like if we had classifiers for software engineers, like holy is that a good compliment?

Okay, cool. Jev is fast at answering basic questions. It's different from an LLM, but do you need it?

Who is it actually for? I feel like a simple answer, which is Jev is for developers. That's the simplest answer.

I think there's like a longer answer that Jev is a little bit for everyone because I think that... AI has been like telling a pretty dark story recently. Take your jobs, kill everyone, etc.

And I think that like we're like back to like early Internet builder energy. And this is my dream. It's better than my dream, to be honest.

Like this part is what makes all of the sleep deprivation worth it. Okay, so clearly Jev is marketed at developers. And anytime before now, I'll be like, oh, well, I'm not a developer, so it's not for me.

But guess what? We're all developers now. All of us are show of hands.

How many of you have generated any amount of code with AI? So we're developers. Jev is for us.

If we're not gonna be talking to Jev, like we talked to Chad, GPT, and it's actually supposed to be talking to our code instead of us. How do we use it? And again, should we?

Yeah, we have a whole new world of creators, of engineers who are either mostly or exclusively using coding agents. They aren't connected to the low -level code. And so if they're not connected to the low -level code, where a primitive fits in is more challenging.

One solution to that is to say, well, I guess we don't actually need... you to understand it. We just need your agent to understand it, which is why we have an official agent skill that we definitely want to, like, improve.

I love what Ali said. Decision models like Jev are a primitive that we don't really need to understand. It's helpful, and I hope you got enough from this video so far.

All we really need to know is that it could potentially help us with our code, our harnesses or whatever. We just need to be able to say, Hey, AI agent, your AI agent, look at this thing and tell me how can it help where in all of our code that we've made or the things we're doing, where could it help us? But I know it's always helpful to see some really cool examples of what people are already doing, like using Jeff and your AI harness to help you make model routing decisions.

So similar to what you might do, I just asked my AI agent, Hey, how can we use Jeff? And one of those was a model router here. I have it set up.

I used open router and Jeff to make some model decisions. Here's a quick demo. Watch it work.

2 plus 2, it used a small model. Write a haiku about coffee, it used medium. Explain BGP, it used medium.

Hold on, it's going fast. Prove there are infinitely many primes. It chose a large one.

It chose Claude Sonnet 5. So what you're seeing is Jeb is making very quick decisions on model routing in moments you don't want to wait. When you ask your question, what is 2 plus 2, you don't want another LLM to sit there and reason like, okay, is this a hard question?

I don't know. And it might cost more money. And then it's so stupid.

But Jeb's like... Easy, boom, easy, hard, hard. Like it's just really good.

Now this one, I'm so excited to show you because it's what I'm using right now. And it's an idea I got from Allie. It's a teleprompter, but instead of it being all of the words that you say, I can't work off a script.

This is how I'm good. I just need to like. just talk right but sometimes there are like bullet points that i need to hit right i need there's like 10 things i need to make sure i cover so the intelliprompter just holds those 10 bullet points does real -time transcription that's using 11 labs because they're really great at that and then it sends that real -time transcription every half second to typesafe to have jev determine has this bullet point been like discussed And for my uses, I only need to know, like, did I start talking about it?

It doesn't have to be even comprehensive. So basically, have I started talking about that bullet point? Good, check it off.

And it just, so this teleprompter is like a dynamic display of what I have yet to talk about. It was going to cost me something like $30 an hour to run. And I wanted to use it for shows.

I do stage shows. So I wanted to have it running for like three hours during a show and not have to worry about it. I'm like, well, this is just like a personal thing.

I can't pay $120 for it to run for three hours. I can't remember the exact costs. So I shelved it.

And then the day I learned about TypeSafe, I actually built it as part of my interview process here, was I revived that project, integrated it with TypeSafe, and suddenly it was like completely reasonable. So I was fascinated by this idea because it's how I record. So I built it with Jeff.

Here it is. It's actually happening right now. It's looking at my script and telling me if...

I've completed parts of my video. Notice I have a live transcript down here and Jeff is quickly telling me and checking off if I've hit parts of my script. So remember this part of the video right here, Jeff only has three answer types.

Yes or no multiple choice and a scoring system. Let's try another one. By the way, this is Diogo Almeida co -founder of type safe.

He actually worked on instruct GPT at open AI before chat GPT. So it's helping me keep track of what I say. And so I don't miss stuff when I'm actually recording, which I do often.

Isn't it an amazing example of Jeff? Like it's checking it in real time. It's quick, but the cost is like nothing.

And by the way, it's not just checking for my exact words. It's checking for the idea. Did I talk about the idea of that?

Cause I, I tend to add live off my scripts. Another demo I saw was putting Jeff into AI web browsing and it made it super fast. So I decided to do one with like ordering a pizza from Domino's.

I'm gonna order a large hand -tossed pizza with pepperoni and mushrooms. Let's see how fast it goes. Ready, set, go.

I noticed it's making the decisions in the terminal right now. I mean, that's it. That was done.

It's pretty fast. Let's try a different order. My favorite type of pizza, a large thin crust pizza.

with Canadian bacon and pineapple. I haven't tested this one, so don't get mad at me if it doesn't work. Ready, set, go.

Dude, it's so fast.

Wow, it did it. I bet you can't order pizza that fast. And that was 15 seconds.

Let's try the same thing with Codex. All right, open the tab. It's doing its thing slowly.

And that took about a minute. And according to Codex, it took about 61 seconds. Now, a few other projects I saw that I loved.

One was Jevgrep. Jev will find the right files for a coding agent, decreasing the time for it to search. The claim is that it reduces your coding agent costs by 40%.

The repo passed 1200 stars in two days. Another is Jev in Hermes. The plugin that will hand yes or no decisions to Jev inside Hermes.

And then one of my favorite ones was the one actually Ally made, Doom. Jev playing Doom. Okay, and I know you're wondering, All these people came out with their own decision models.

Jeff is supposed to be special, but now we have cleft from. Cloudflare. I almost forgot.

That's a local model. I'm actually running that locally here. We've got the decisions API that's from open AI.

They just announced that at their dev day recently, I'm running a different size clef, and then I'm running nimble and Tev, which is from a llama right now. Also running locally on my Mac. And what I'm gonna do is run this on 20 real bot comments, 107 real fan comments and see what they can classify it as ready set.

I'm excited for this. I haven't seen it yet. Go.

This is fun. Woo. All right.

27 B is taking a minute. Okay. Here's the results.

Jev clear winner decisions API from open AI coming in at a close second. Everything else is local and it's not bad. It also looks like accuracy goes to our cloud models and most accurate on the local side was the cleft 27 B Jev is still the cheaper one.

As far as cloud goes open AI's decision API is 2 .4 times more expensive. And on these tests, it looks like Jev is more accurate, a tad bit faster, but they're kind of neck and neck that's Jeff, this new thing, this new decision model, this new class of model system one that everyone was kind of freaking out about.

And they still are. I find it fascinating that the devs are getting really excited about this. When I first heard about it, I'm like, okay, I don't get it.

And it took me a minute to die then and actually kind of appreciate what it is and talking with Diogo and talking with Ali and Just seeing the reaction on Twitter made me realize, oh, Twitter, X, made me realize this is actually something pretty cool. This might be a new class of model that will enable us to do some really cool things.

I love what Allie said about it, comparing it to Docker. I'll let her say it here. I like to draw a parallel to Docker and AWS Lambda, serverless function execution.

Both of these technologies, when they were new, they were really hard to like... wrap your head around. You can't imagine being in any sort of infrastructure without having like a firm grasp of like Docker.

Docker is now like the main thing we use. Like you couldn't imagine a world where developers aren't hosting their apps on Docker, but that's kind of how it was back then when Docker was introduced. And I had a similar experience.

I'm like, Docker, what is this? And to like understand that and explain it, it took me a minute. Decision models like Jev might be in that moment right now until we all become aware of like, oh, this is what it can do.

Now I have ideas. I could do this, this, and this like an IntelliPrompter. That was just one random idea I got from alley.

And I'm already like, this is pretty sick. I'm gonna use it anyway. That's all I got.

Let me know in the comments below. Like, are you going to use Jeff? Do you have any ideas for things like Jeff?

And also have you hacked the YouTube algorithm today? You gotta make sure you do. You gotta hit like comment, subscribe notification bell.

You gotta hack YouTube today. Ethically. Of course.

All right. That's all I got. I'll catch you guys next time.

Hey, You made it to the end of the video. At the end of my videos, I like to pray for you, my audience.

It's kind of weird. It's okay. Just go with it.

You're not here by accident. I mean, how often do you make it to the end of a YouTube video? So here we go.

One, two, three. Pray. God, I thank you for the person on the other side of the screen, on the other side of this camera.

I thank you for who they are. I ask right now that you bless them indeed right now. Bless them and their family.

Fill them with joy and peace. You know, this video is about decisions, about using AI to help make decisions, but also we as humans have to make some very important and often life -changing decisions. So I ask right now that this person I'm praying for, if they're facing a big decision right now, that you would give them wisdom, that you would give them insight, that you would allow them to commit their way to you and to plan their way but trust you to direct your steps or direct their steps.

Just give them wisdom in that decision, God. And some decisions are really hard and rough and deep and others are like, oh, this could be a great opportunity. I don't know.

So God, just give them wisdom in that. Help them to walk into it with confidence. And I just thank you for who this person is.

Fill them with joy and peace right now. Fill them with excitement. And I pray ultimately this person can know You, Jesus.

I ask this in your name, Lord. Amen. Thank you for letting me do that.

I'll catch you guys next time.

The Hook

The bait, then the rug-pull.

Jev can't talk, can't search and can't call tools, and developers are copying it anyway. NetworkChuck runs it head to head with ChatGPT and finds that giving up conversation is exactly what makes it useful.

Frameworks