🚀 OpenAI launches all-in-one agent

Welcome back, Superhuman. It’s not GPT-5, but it’s awfully close. OpenAI’s new ChatGPT Agent blends the best of its previous releases into one general-purpose tool, getting us that much closer to a truly all-in-one AI assistant. And: Voice models take another leap toward realism.

Also Today: Learn how to build interactive tools and apps without code, and get the latest prompts, headlines, and social posts.

FROM THE FRONTIER

OpenAI stuns with surprise all-in-one agent

Sam Altman is “feeling the AGI” — and for good reason: OpenAI just launched a feature called ChatGPT Agent that can seamlessly transition between “thinking” and taking action, essentially combining ChatGPT, deep research, and Operator into a single tool.

How does it work? The agent can use its own virtual computer and share its chain-of-thought as it decides what to do next. It can also access a terminal, letting it analyze and run code, generate files, and connect to your private data sources. You can interrupt at any time to add more instructions or clarify something.

Some examples:

• Ask ChatGPT to analyze a wedding invitation to help you find an appropriate outfit, brainstorm some gift ideas, book your hotel, etc.

• Upload a photo, then ask ChatGPT to turn it into swag for an upcoming launch; you can get pretty specific, e.g., “order 500 1×1 laptop stickers and send them to X address”

• Use it to generate a slideshow, complete with charts and visuals; it’ll even evaluate its own slides and fix ones that aren’t up to par

• Click here to watch several more real-world examples in action

How does it perform? On the notoriously difficult Humanity’s Last Exam, it scores a state-of-the-art 41.6, beating deep research by 15 points. It gets similarly impressive results on FrontierMath, BrowseComp, and DSBench.

What are the risks? To prevent prompt injection attacks, the agent has built-in monitors that can flag when a website looks suspicious. But OpenAI admits this system isn’t foolproof, and Agent should be used with caution.

When can I use it? It’s already started rolling out to subscribers. Pro users get 400 queries per month, while Plus and Team users get 40.

PRESENTED BY SYNCLY

All your customer feedback—aggregated in one place

Tired of tracking customer sentiment in a dozen places? Not anymore.

Syncly gathers all your feedback—support tickets, reviews, social comments—into one clear view, so you can finally see the full picture.

Grow faster with the customer insight you need:

• Auto-tags, organizes, and surfaces real insights fast

• Search data using ChatGPT-like prompts

• Spot issues instantly with real-time alerts and automated dashboards

Book a demo using your own data + enjoy a 7-day trial at no cost.

TODAY IN AI

1. Hume’s new voice model can mimic your personality: Hume’s EVI 2 was already remarkably realistic, with a focus on the expressive, emotional qualities that make every voice unique. Now it’s opening up access to EVI 3, a model that doesn’t just clone your voice but captures more of your personality, down to your unique speaking style and vocabulary. It’s great for building language coaches, podcast hosts, storybook narrators, and more. You can try it here.

2. Vibe coding startup Lovable catapults to unicorn status: Despite being less than a year old, Swedish AI startup Lovable just landed $200M in a Series A round that values the app-building platform at $1.8B. With 2.3M active users, the company has made a name for itself by catering to non-coders, including many who use it to draft mockups and prototypes before bringing them to developers for additional help.

3. Adobe revamps Firefly video with DIY sound effects: Who knew video editing could be so fun? Now, instead of relying on pre-rendered SFX to accompany your clips, you can record your own voice (e.g., “Boom!” for an explosion or “crunch, crunch” for snapping tree limbs), then use text prompts to turn it into a custom sound effect. Here’s a fun example of how it works from creator KarenXCheng. Other Firefly updates include advanced composition controls and access to more third-party VLMs.

THE AI ACADEMY

How to build interactive tools and apps without code

• Go to Super Grok and sign up.

• Enter your prompt describing your idea in detail.

Sample Prompt: “You're a senior software architect skilled in no-code/low-code platforms and web tech.

Task: Build an MVP tool based on this idea: [Insert your tool idea]

Deliverables:

• What it does: Simple explanation of the tool and who it’s for.

• Build plan: Step-by-step using a no-code/low-code platform.

• Code (if needed): HTML/CSS/JS for key parts.

• Design tips: Make it simple, clear, and user-friendly.

• How to share: Steps to publish/embed the tool.

Imagine you are shipping an MVP for a startup demo.”

• Wait for a few minutes and Grok scaffolds the UI, logic, and deployment steps , no coding required.

PRESENTED BY STACKAI

Finally, the no-code agent builder we’ve been waiting for

With StackAI, you can deploy AI agents across your business fast—no code, no friction, no risk.

More flexible than Glean. More scalable than n8n.

• Built-in support for LLMs, function calling, and RAG

• 100+ integrations like SharePoint and SAP

• Security compliance you need: SOC 2, HIPAA, etc

Try StackAI today.

IN THE KNOW

What’s trending on socials and headlines today

🎥 Movie Magic: a16z’s Justine Moore shows how to create an AI-generated video set in a real-world location (for example, a stormtrooper vlogging inside the a16z office).

💰 Cashing In: OpenAI research scientist Hyung Won Chung – who is allegedly joining Meta soon – gave a 30-minute master class on how to create wealth in our new AI-powered economy, racking up at least 1M views.

📱 Watch and Learn: Builder Steven Cravotta shows how to use AI to generate your own viral app, including the best ways to come up with unique ideas.

✨ Also Worth Checking Out: An LLM chess showdown, 10 ways to use Perplexity’s Comet browser, and video game concepts made with AI.

Slack announced a new suite of AI features, including summarization tools, meeting transcriptions, and even message explanations to help you decode acronyms and company jargon.

Google just added its Veo 3 model to the Gemini API, letting developers build their own apps with the cutting-edge VLM.

Mistral is rolling out major updates to its consumer-facing chatbot le Chat, including Deep Research, a voice mode, a Projects feature, and advanced image editing. Here’s a video with more details.

PRODUCTIVITY

5 New AI Tools to Boost Your Productivity

• 🔍️ Seamless.AI*: Find & connect instantly with 1.3B+ decision-makers with verified insights, like contact info, buyer intent, company intel, and more. Sign up today.

• 🏦 Modernbanc: Chat with your business finances and get answers instantly.

• ✅ Memno: Remembers, organizes, and handles everyday tasks for you.

• 💻️ Mysite: Build websites for small businesses without coding skills.

• ⚙️ Cekura: Automate QA, testing, and observability for conversational AI voice agents.

^{* indicates a promoted tool, if any}

PROMPT STATION

Visualize data with ChatGPT

[upload data file to ChatGPT]

Role: You are an expert data analyst and data visualizer.

Context: I have attached a file about [describe file]

Task: Analyze the data in the file. Generate a list of key trends and insights. Visualize key trends and insights with relevant charts and data visualizations.

Note: please be careful when uploading sensitive company or personal data to ChatGPT.

Real or Rendered?

Whenever you’re ready to take the next step

• Check out our Top 125 AI Tools

• Get better at prompting with our Top 1,000 Prompts

• Learn how to use AI at work and get certified with our free course

Grow customers & revenue: Join companies like Amazon, Hubspot, and Salesforce. Showcase your product to our 1M+ readers and 2M+ followers on socials. Get in touch.

Until next time — Zain & the Superhuman AI team