🍭 Anthropic's New Models Any Good? 🤔

Good morning. While AI models keep dropping we keep asking? They any good?

Meanwhile our intern keeps avoiding building systems. He’ll change his mind after reading this.

Let’s dig in.

What’s Cookin’:

• Claude 4, Opus 4 - Deep dive

• How Tiny Structures Create Big Confidence

• 6 Tools to help you schedule, reply, and get more done!

• Steal This Prompt: Make it Felt

• And Everything Else you Need to Know

Anthropic
🤔 Two New AI Models. One Question.

The Bite: While Claude 4 shows strength in coding and reasoning tasks, it lags behind in multimodality (images, audio, video) and context window size.

We dropped the overview of Anthropic’s announcements on Monday.

Let’s dive a bit deeper into the two new models, Opus 4 and Claude 4. 

Where they stack up against the comp, benchmarks, and what this means going forward.

And answer the single question: r they gud?

Snacks: 

• Both new models have a 200K token context window. Far below the industry leaders.

• Just a year or two ago, context windows of 8K, 32K, or 128K were considered cutting-edge.

• Both models: designed to power more capable, autonomous AI agents for multi-step workflows

• Opus 4: Built to handle long tasks, deep research, and big-brain workflows.

• Opus 4 touted as the “worlds best coding model” 

• Opus 4 (72.5%) and Sonnet 4 (72.7%) models demonstrate leading performance on the SWE-bench for coding

Why it bites: Despite some criticism, which we think is mostly sensationalism, they are building something amazing and it’s hyper-intentional.

You mad they don’t have image and video gen?
You aren’t their customer.

Yes, it’s still possible for a multi billion dollar company to niche down.
And niche down again.

Are these new models good? Yes
Do they have a smaller context window than a handful of LLMs. Yes

Is there a reason your favorite vibe coder on YouTube still was using Claude 3.5 even after other models dropped with higher benchmarks and larger context windows? Yes

The reason: They figured out what their model could be good at one thing.
Coding.

They niched down.
Then niched down again.

OpenAI/Google -> Swiss Army Knife
Anthropic -> Brain Surgery Tool 

🍭 Steal this Prompt
🧸 Transform Anything Into Felt

Ever wanted to make anything you can think of a soft felt version of it?

Me neither.
But now you can!

1. Think of of the #object you want to remix

2. Hit this link (prompt + references)

3. Once in GPT 4o, replace the “#object” with your subject

4. Watch it turn your subject full cozy.

🍭 Try the prompt on Snack Prompt 🍭

Big Biz
👻 Invisible Systems

How Tiny Structures Create Big Confidence

The Bite: Growth for solopreneurs usually brings chaos first, structure second.

But what if it didn’t have to?

What if you had tiny, almost invisible structures in place from day one—so growth was exciting, not anxiety-inducing?

Eder calls this “building the foundation correctly”—spending your effort on the prep so execution becomes frictionless.

Snacks: 

🍭 Real Confidence = Invisible Control
→ Businesses fear adopting AI because they worry about losing control.
→ Small structures (like a KMS & OMS) give solopreneurs invisible confidence.
→ You build it once, then trust it forever.

🍭 Structure Isn’t Bureaucracy. It’s Certainty
→ People choose what’s certain over what’s lucrative. Tiny, reliable systems create clarity without complexity, letting solos keep moving fast.

🍭 The Prep-Work Principle
→ Eder told us a story about painters today in our community call.
→ Investing time upfront (like prepping the room carefully with tape) drastically accelerates execution.

Small, lightweight structures let you move faster later because you’re not constantly second-guessing.

Why it Bites: We recently watched a solo automation expert quietly land a huge client—but this wasn’t luck.

They’d done the invisible prep-work:
→ Set up a "context layer”
→ Built a simple "validation layer”

Suddenly, onboarding and execution became simple, predictable, even boring (in a good way).

They didn’t slow down to “align” or “get permission.”

They just shipped confidently—because the tiny structure they built once now gave them invisible leverage every time.

The big client never saw the systems behind the curtain.

They just saw results—fast and consistently high-quality.

That’s what stupid simple systems do:
→ They transform growth from something terrifying into something quietly thrilling.
→ And they create a confidence that scales, without adding friction.

Together with Taplio
Grow your LinkedIn audience with the right AI

Taplio’s AI is trained on 500M+ LinkedIn posts to help you write high-performing content in seconds.

✅ Pick your hook & format;

✅ Generate a full post with AI;

✅ Edit, refine, and schedule—all in one place.

“Taplio keeps me inspired and makes posting effortless. It’s the most complete tool for anyone serious about LinkedIn growth." — Justin Welsh, 700K followers & $9.1M generated on LinkedIn

Use the code "DAILY1X1" to get one month for $1!

Tool Box 
🧰 Trending AI Tools

📥 Zero — AI-native email client that manages your inbox for you

💬 Sociable AI — predicts viral posts and drafts top social comments + replies to DMs

🐰 CodeRabbit — Context-aware AI-powered code reviewer

→ Automated code reviews with real project context. Catches bugs, flags security issues, and suggests clean refactors.

Works with GitHub, GitLab, VS Code — and yes, it actually saves you time.

📊 Crowdwave — run fast, scalable surveys using simulated respondents for instant insights

📊 Skywork Super Agents — turns input into docs, slides, sheets, podcasts, and webpages

📞 SchedX — AI SDR that talks to site visitors, qualifies leads, and books meetings

Everything Else
🧠 You Need to Know

📰 AI Agents are making morning news!

→ Micro Center News editor Dan Ackerman joins "CBS Mornings Plus" to share his test. Creating his own AI "assistbot" to manage real tasks like scheduling, vacation planning and staying organized.

🎈 AI’s carbon footprint is ballooning — data centers are guzzling power faster than grids can green

→ EPA may drop emission limits just as AI energy use explodes. Great timing.

🚓 Cybercriminals love AI — deepfake scams, stolen logins, poisoned models, and phishing at scale

→ One British firm lost £20M on a fake Zoom call. Yes, really.

🤫 AI agents are leaking secrets at scale — one bot = 45 credentials, zero adult supervision

→ GitHub leaked 23M secrets last year, and Copilot made it worse by 40%.

🦺 Amazon devs say coding feels like warehouse work now — same output, half the team, all the speed-up

→ Some engineers call it “Copilot or die.” Others just miss thinking.

Referral Program
Invite Friends & Earn Some Freebies 🔍