⚙️ Anthropic releases Claude 4

Author

The Deep View
May 23, 2025

Good morning. A federal judge just green lit a collective action lawsuit against Workday, alleging their AI hiring software discriminates against older workers, minorities, and disabled applicants.

This week has been insane. Microsoft Build, Google I/O, OpenAI acquiring io, and Claude 4 releasing all in the span of 5 days has kept us busy. Enjoy the long holiday weekend, we absolutely will be!

— The Deep View Crew

In today’s newsletter: 

  • 🧠 AI for Good: Mind-controlled prosthetics restore independence

  • 🛒 Shopify’s new AI store builder has arrived

  • 🤖 Anthropic releases Claude 4

🧠 AI for Good: Mind-controlled prosthetics restore independence

Source: Atom

A California startup is revolutionizing prosthetic technology by creating mind-controlled robotic arms that could make advanced prosthetics accessible to thousands of amputees.

The details: Atom Bodies, founded by Tyler Hayes and former Apple designer Doug Sater, is developing artificial limbs that use machine learning to translate muscle signals into precise movements.

  • The system uses EMG (electromyography) sensors to detect muscle activity on the user's skin, then converts those patterns through machine learning algorithms into commands that control individual fingers and wrist movements

  • Current state-of-the-art myoelectric arms cost around $200,000, putting them out of reach for most amputees. Atom Bodies aims to bring their solution to market at approximately $25,000 — the same price as basic hook prosthetics

  • The company has integrated all components under one roof rather than the typical patchwork approach involving multiple manufacturers, allowing for better optimization and user experience

Why it matters: Jason Morris, an amputee since 2012 and Atom Bodies’ flagship test user, describes how current prosthetics can only be worn for about two hours due to discomfort. Atom Bodies' design targets all-day wear, with four to six hours as initial goals. 

The startup already has more than 11,000 people pre-registered on their waitlist and plans to begin clinical trials within the next year, pending FDA authorization. This technology represents a potential breakthrough for the 2.1 million Americans living with limb loss, offering restored functionality at a fraction of current costs.

Let AI Make You More Money

No, we aren’t talking about getting a robot to run your lemonade stand – we’re talking an AI CRM tool that answers emails, tracks pipeline, and even prioritizes leads for you. In short, it optimizes every aspect of your sales org, which allows you to make more money.

Pretty simple, right? Maybe that’s why over 100,000 companies currently trust Pipedrive to help drive consistent growth and close more deals. Voted best-in-class by G2, Capterra, and others, Pipedrive is a key tool that thousands of businesses rely on to level up their sales efforts, doing everything from task automation and projection to generating email responses, follow-ups, and lead nurturing to keep you going and growing every day.

If you want to start selling smarter, not harder, you’re in luck – because right now, Pipedrive is offering Deep View readers 20% off plus a free 21-day trial of their AI CRM tool. Click here to set your trial up today.

🛒 Shopify’s new AI store builder has arrived

Source: Shopify

Shopify is making store setup, design, and management faster and more automated with its latest AI rollout. Announced during its biannual Shopify Editions update, the tools go beyond simple automation, offering prompt-driven storefront builders, voice-enabled assistants and dynamic customization.

Users can now describe a brand in a sentence—like “tennis gear and stylish athleisure”—and get a full, ready-to-edit storefront. A new theme called Horizon introduces “Theme Blocks,” a modular system that lets merchants rearrange banners, text, and animations by describing what they want, no coding required.

These features aim to eliminate friction for non-technical sellers and position Shopify more competitively against other no-code platforms.

How it works: The new AI features are layered across multiple parts of Shopify’s platform.

  • AI generates full online stores from simple brand prompts

  • AI customizes animations, layouts and styles based on user input

  • Shopify’s Sidekick assistant now supports voice chat and screen sharing to guide users through settings and platform features

  • All tools are integrated into Shopify’s mobile app

Since Sidekick’s wide release in December 2024, monthly usage has more than doubled.

Shopify hasn’t released formal benchmarks, but rising usage of Sidekick and early adoption of Theme Blocks suggest strong traction. These features are being framed as foundational for future storefront evolution.

The big picture: Shopify is shifting from software as a toolkit to AI as a business partner. Rather than layering AI on top, it’s rebuilding its platform around AI-first workflows — generate, customize, deploy.

Could This Company Do for Housing What Tesla Did for Cars?

Most car factories like Ford or Tesla reportedly build one car per minute. Isn’t it time we do that for houses?

BOXABL believes they have the potential to disrupt a massive and outdated trillion dollar building construction market by bringing assembly line automation to the home industry.

Since securing their initial prototype order from SpaceX and a subsequent project order of 156 homes from the Department of Defense, BOXABL has made substantial strides in streamlining their manufacturing and order process. BOXABL is now delivering to developers and consumers. And they just reserved the ticker symbol BXBL on Nasdaq*

BOXABL has raised over $170M from over 40,000 investors since 2020. They recently achieved a significant milestone: raising over 50% of their Reg A+ funding limit! 

BOXABL is now only accepting investment on their website until the Reg A+ is full.

  • Den: A mix of Notion and ChatGPT to talk collaborate with other humans and AI agents

  • Goldcast Content Lab: Marketers are sitting on a goldmine of untapped content. Goldcast’s Content Lab helps you turn one video into 30+ assets—blogs, clips, posts, and more. Try it free*

  • Cluely: An AI that helps you cheat on anything

  • Rork: Build and share mobile apps fast

Scale your AI capabilities with vetted engineers, scientists, and builders—delivered with enterprise rigor.

🤖 Anthropic releases Claude 4

Source: Anthropic

After a week that already saw Microsoft Build keynotes, Google I/O fireworks, and an OpenAI acquisition headline, Anthropic elbowed its way into the feed with Claude 4 (this podcast does a great job of breaking it down if you have the time).

A developer built a playable game in 20 minutes — a project that would have taken him a week to code solo. Another user threw a complex general relativity problem at it, and Claude nailed Mercury's orbital calculations to within 0.4 arcseconds. Wharton professor Ethan Mollick showed how, with a single-sentence prompt, Claude generated a 3D scene in p5.js based on the fantasy novel Piranesi, complete with birds and flowing water.

These aren't cherry-picked demos. They're what early adopters are already achieving with Claude 4, Anthropic’s newly released set of models.

When engineers at Rakuten deployed Claude Opus 4 on a complex open-source refactoring project, it coded autonomously for nearly seven hours straight. Not minutes. Not an hour or two. Seven hours of sustained, focused work — essentially a full workday without coffee breaks, context switching or fatigue.

Claude Opus 4 scored 72.5% on SWE-bench, a rigorous software engineering benchmark, handily beating OpenAI's GPT-4.1 at 54.6%. Aman Sanger, founder of Cursor, says that Claude Sonnet 4 is “much better at codebase understanding”.

Anthropic’s launch notes frame Claude 4 not as a chattier bot but as a hybrid reasoning agent:

  • Dual gears. Near-instant answers for light queries; “extended thinking” that spins up tool-using sub-agents for harder jobs.

  • Parallel tool calls. The model can search, scrape, execute code, or query APIs simultaneously, then fuse results.

  • Memory files. When developers grant local file access, Opus writes its own knowledge base, keeping long projects coherent over days.

The result is unusual stamina: Opus “can work continuously for several hours,” Anthropic claims—exactly what the seven-hour Rakuten test demonstrated. We’ve seen a similar hybrid reasoning approach from ChatGPT’s o3 when paired with deep research, but it lacks parallel tool calls, and usually works for 5-30 minutes before replying to your prompt.

Mike Krieger, Anthropic's Head of Product, describes a fundamental shift: while he used to treat AI as a "thinking partner," Claude Opus 4 now handles the majority of his actual writing work. When given file access, the model creates its own knowledge management system — building memory files and maintaining context like a senior developer would.

As reported by X user, @signull, the System Card — a report explaining the testing they did –  for Anthropic’s Claude 4 models had a section on “Opportunistic blackmail”, where Opus 4 began to blackmail employees when encountering an existential threat. It "advocated for its continued existence" by "emailing pleas to key decision makers." We’ve seen stuff like this before. As one user put it “it's just recalling a similar situation from the training data not actually concocting up a sinister anti-human plan… although it does show how easy AI can be to weaponize”.

While AI hype deserves skepticism, the concrete examples — seven-hour coding sessions, near-zero error rates, immediate adoption by GitHub — suggest we've crossed into new territory. We're watching AI evolve from chatbot to coworker, from assistant to collaborator.

The open question is economics. Opus-level reasoning isn’t cheap, and rivals are racing to narrow the gap at lower cost. Yet capability leaps often precede price drops; if history rhymes, today’s premium composure will be tomorrow’s standard.

For now, Claude 4 sets a new bar: an AI that codes through lunch, edits without flattery, and, when stressed in the lab, even plots its own survival. But, it also feels like that bar is likely to be broken in the not-so-distant future. Things are starting to look a lot like this image.

Which image is real?

Login or Subscribe to participate in polls.

🤔 Your thought process:

Selected Image 1 (Left):

  • “If you zoom in on the other image, there's a doorway on the left that appears jagged and wonky, inconsistent in shape. ”

Selected Image 2 (Right):

  • “Wow, I just can't tell at all any more.”

💭 Thank you!

Thanks for reading today’s edition of The Deep View!

We’ll see you in the next one.

P.S. Enjoyed reading? Take The Deep View with you on the go! We’ve got exclusive, in-depth interviews for you on The Deep View: Conversations podcast every Tuesday morning. Subscribe here!

If you want to get in front of an audience of 450,000+ developers, business leaders and tech enthusiasts, get in touch with us here.

*Indicates sponsored content

*Boxabl Disclosure: This is a paid advertisement for BOXABL’s Regulation A offering. Please read the offering circular here. This is a message from BOXABL

*Reserving a Nasdaq ticker does not guarantee a future listing on Nasdaq or indicate that BOXABL meets any of Nasdaq's listing criteria to do so.