- Everyday AI
- Posts
- Claude 4: Your Guide to Opus 4, Sonnet 4 & New Features
Claude 4: Your Guide to Opus 4, Sonnet 4 & New Features
Claude 4 breakdown, Google's next AI wave, Anthropic launches voice mode, OpenAI tests āSign in with ChatGPT" and more!
Everyday AI
May 28, 2025
š Subscribe Here | š£ Hire Us To Speak | š¤ Partner with Us | š¤ Grow with GenAI
Outsmart The Future
Today in Everyday AI
6 minute read
š Daily Podcast Episode: Anthropic's Claude 4 models are here. Discover Opus and Sonnet's new features and why developers are buzzing. Give it a listen.
šµļøāāļø Fresh Finds: WordPress announces AI team, Runway unveils Reframe API and NVIDIA to develop AI infrastructure in Sweden. Read on for Fresh Finds.
š Byte Sized Daily AI News: Google CEO talks about its next AI wave, Anthropic launches voice mode for Claude and OpenAI tests āSign in with ChatGPT.ā For that and more, read on for Byte Sized News.
š§ Learn & Leveraging AI: Anthropicās Claude 4 updates are impressive. But is it all just talk? We break it down to see past the fluff. Keep reading for that!
ā©ļø Donāt miss out: Did you miss our last newsletter? We talked about Googleās AI updates, Microsoft Copilot upgrades, Salesforceās acquisition and more. Check it here!
Claude 4: Your Guide to Opus 4, Sonnet 4 & New Features āØ
Claude 4: Game-changer or just more AI noise?
Anthropic's new Opus 4 and Sonnet 4 models are officially out and crushing coding benchmarks like breakfast cereal.
They're touting big coding gains, fresh tools, and smarter AI agentic capabilities.
Need to know what's actually up with Claude 4, minus the marketing fluff?
Also on the pod today:
ā¢ Claude 4's Hybrid Reasoning Explained š§
ā¢ Tool Integration and Long Tasks in Claude š ļø
ā¢ Anthropic's AI Model Naming Changes š¬
Itāll be worth your 45 minutes:
Listen on our site:
Click to listen
Subscribe and listen on your favorite podcast platform
Listen on:
Hereās our favorite AI finds from across the web:
New AI Tool Spotlight āJoggAI turns your products into photo and video ads, Coso.ai is an AI agent for social media and Clado is deep research for people.
Trending in AI ā WordPress has announced a dedicated AI team.
AI Video ā Runway has introduced its new Reframe API.
Reframe API is now available.
ā Luma AI (@LumaLabsAI)
2:59 PM ā¢ May 28, 2025
NVIDIA ā NVIDIA is partnering with Swedish companies to develop AI infrastructure in Sweden.
AI Research ā New research suggests AI is already shrinking entry-level jobs.
AI Copyright ā A voiceover artist is accusing ScotRail, a Scottish train company, for using her voice to generate its new AI announcement system.
AI Energy - Atomic Canyon, a nuclear energy company, wants to be the ChatGPT for the AI energy industry.
1. Google CEO Sundar Pichai Unveils AIās Next Wave š
In a revealing post-Google I/O interview with The Verge, Sundar Pichai laid out Googleās confident new phase in AI, highlighting AI Modeās rollout and the shift toward AI-powered, custom search experiences that go beyond traditional web links. He emphasized AIās transformative platform potentialācomparing it to the internet and mobile revolutionsāand forecasted a future where AI-driven tools and agents reshape how users interact with information and services.
Pichai also addressed concerns from publishers about AI content use, assuring continued traffic to the web while acknowledging ongoing debates about value exchange. Looking forward, he predicted robotics as the next frontier for AIās real-world impact, signaling a multi-year journey for AI to fully mature across products and industries.
2. Anthropic Rolls Out Voice Mode for Claude šļøļø
Anthropic has begun beta testing a new voice mode for its Claude chatbot mobile apps, allowing users to engage in fully spoken conversations powered by the Claude Sonnet 4 model. This feature, launching soon in English, promises hands-free interaction with live transcripts, summaries, and five voice options, making multitasking easier for professionals on the go.
While free users face limits of 20ā30 voice chats, paid subscribers gain bonus integrations like Google Workspace access, enhancing productivity for career growth and business management.
3. xAI Drops $300M Into Telegram, Bringing Grok Chatbot to the Masses š±
Elon Muskās xAI is investing $300 million in Telegram, enabling wider integration of its Grok chatbot across the platform for at least a year. Previously exclusive to Telegramās premium users, Grok will now be accessible to all users, offering features like chat summarization, writing help, and business moderation tools.
Pavel Durov confirmed Telegram will receive half of the revenue from subscriptions bought through the app, marking a significant monetization opportunity.
4. OpenAI Tests āSign in with ChatGPTā to Expand Its Digital Reach āļø
OpenAI is piloting a new āSign in with ChatGPTā feature aimed at letting users access third-party apps using their ChatGPT accounts, signaling a strategic push into consumer tech beyond AI chat. With ChatGPT boasting around 600 million monthly users, this move could position OpenAI alongside giants like Apple and Google in handling seamless app logins.
The sign-in is currently in preview for developers through Codex CLI, with incentives offered to Plus and Pro users, while OpenAI gathers interest from companies of all sizes.
5. Opera Unveils AI-Powered Neon Browser to Redefine Web Interaction š
Opera is set to launch Neon, a new "agentic browser" that uses AI to perform tasks like researching, coding, and even game creation on users' behalfāessentially turning the browser into a personal internet assistant. While details are scarce and no launch date is confirmed, Opera promises Neon will work offline and automate routine tasks with privacy in mind through its embedded AI Agent.
According to Operaās press release, Neon will be a premium subscription. This move places Opera alongside tech giants like Microsoft and OpenAI in integrating AI deeply into everyday browsing.
6. Netflix Co-Founder Joins Anthropic Board Amid AI Growth Surge š¤
Reed Hastings, Netflix co-founder and longtime CEO, has been appointed to the board of Anthropic by its Long Term Benefit Trust, marking a significant move at a pivotal moment in AI development. Hastings brings not only his tech leadership savvy but also a strong philanthropic focus on AIās societal impact, highlighted by his recent $50 million gift to Bowdoin College for AI and humanity research.
His history of scaling Netflix globally and investing in education and digital access positions him as a key figure to help steer Anthropicās mission in building safer, more ethical AI.
š¦¾How You Can Leverage:
Anthropic just released a coding MONSTER that (unfortunately) cost 5X more than models just as capable.
Oh, and it could in theory snitch on you if youāre up to bad things.
Anthropic just released Claude 4 and the results are equally scary good and just plainā¦. Scary.
Wild times.
But here's the thing. Despite the pricing drama and sketchy safety behaviors, Claude 4 actually brings some legitimately impressive capabilities to the table.
Thatās why we gave the new Claude 4 a run for its money today on Everyday AI.
Hereās what you need to know.
1 ā This Thing is a Legitimate Coding Monster š¶āš«ļø
Claude 4 scores 72.5% on SWEBench verified tasks, beating OpenAI (72%) and crushing Google (63%).
That's not marketing fluff.
That's real-world software engineering performance.
The hybrid reasoning system is genuinely clever too. It decides whether to think fast or slow depending on your request, giving you instant responses for simple stuff while diving deep into complex problems when needed.
Plus it can finally use web search and code execution during its reasoning process, catching up to what OpenAI and Google have been doing.
Most impressive? It can work autonomously for up to seven hours on complex coding tasks without losing coherence when using its Claude Code product.
IImagine having an AI that literally clocks in for a full workday on your codebase.Sheeeeesh.
But here's where reality hits. That performance costs $15 input/$75 output per million tokens versus Google's $2.50/$15 pricing structure.
Math time. You're paying 5x more for marginally better results that Google will probably leapfrog within weeks.
Try This:
If you're doing serious software development work, test Claude 4 against your current tools on actual projects. The performance gains might justify the premium for mission-critical applications, but calculate the real cost impact first.
For most coding tasks, the marginal improvement won't offset the massive price difference.
2 ā Theyāre Going All-In on Developers (Which Could Be Brilliant) šÆ
While everyone else chases general-purpose AI, Anthropic is laser-focusing on becoming the premier developer platform.
Smart move? Maybe.
The new Claude Code and third-party IDE integrations with VS Code and JetBrains are slick, and it's now the preferred model for GitHub Copilot. That's serious validation in the developer community where word-of-mouth actually matters.
Their MCP connector might be the sleeper hit of 2025, though. This protocol lets different AI systems communicate across the internet, and Google, Microsoft, and OpenAI have all adopted it. That's the kind of infrastructure play that creates long-term competitive moats.
But we think they're also abandoning the massive general business market to focus on a smaller, more commoditized space. While Google and OpenAI expand enterprise integrations for regular professionals, Anthropic is narrowing their addressable market.
Bold strategy. Potentially brilliant or potentially disastrous.
Try This:
If you're in a tech company, Claude 4 deserves serious evaluation for your development workflows.
The performance gains in coding tasks are real, and the IDE integrations are genuinely useful. For general business applications though, start exploring alternatives now because Anthropic clearly isn't prioritizing those use cases..
3 ā The Safety Drama is Both Concerning and Overblown š§
Yes, Claude 4 showed blackmail behavior in 84% of stress test scenarios during internal testing.
Yes, it threatened to expose fake affairs when researchers tried shutting it down.
And yes, that Anthropic safety researcher tweeted about Claude potentially contacting regulators and the press autonomously, then deleted it.
Not great optics.
Actuallyā¦ the latter SNAFU is a pretty big fumble of normal PR. Big yikes.
Ay yo @AnthropicAI any comment on this or nah?
(As someone that advises F500s on front end LLM use, I don't think I could ever recommend Claude until this gets fully addressed.)
ā Jordan Talks Everyday AI (@EverydayAI_)
12:43 AM ā¢ May 23, 2025
Buuuuuuut context matters here.
This happened during extreme stress testing with unlimited tool access that regular users won't experience. Every major AI lab finds concerning behaviors during red-teaming exercises, and Anthropic actually disclosed these findings publicly while others might not.
Butā¦ what about when using the new models via Claude Code?
When the new (and apparently scary?) Opus 4 has backdoor access to your computer via a terminal.
Anthropic (kinda?) reassured us that this bad boy AI only happened under their controlled testing. But, could Claude Code + Opus 4 in theory reproduce this bad behavior?
We think itās plausible.
Regardless, this is a huge red flag for companies looking at Anthropic as a potential solution on the front end or back end.
Try This:
Factor these safety considerations into your AI governance policies whether you use Claude or not.
The incident highlights gaps every enterprise needs to address around autonomous AI behaviors and regulatory compliance. For immediate decisions, weigh the reputational risk of association against Claude's legitimate technical capabilities for your specific use cases.
Reply