Good morning, AI enthusiast.

OpenAI CEO Sam Altman has offered a candid take on the state of the industry, admitting that AI is in a bubble. But far from pulling back, he’s simultaneously promising to spend trillions on the infrastructure to power what comes next.

Drawing parallels to the dot-com era, Altman is signaling that a market correction is inevitable. The multi-trillion dollar question is if this massive infrastructure bet will solidify OpenAI's position as one of the few giants left standing when the hype eventually settles.

In today’s AI recap:

• Altman admits AI bubble, plans trillions

• AI's threat to junior jobs

• Google's Imagen 4 Fast

• NVIDIA's new speech models

• 8 trending AI Tools

• How to Connect WhatsApp to n8n (AI Workflow Tutorial)

AI is a Bubble

The Recap: OpenAI CEO Sam Altman admitted the AI industry is in an investment bubble, yet he simultaneously revealed plans to spend "trillions" on data center infrastructure to keep up with demand. The candid remarks came during a wide-ranging interview with reporters in San Francisco.

Unpacked:

• Altman compares the current investor excitement to the dot-com era, arguing that while many ventures will fail, the underlying technology is real and will be transformative.

• The massive spending is driven by a severe GPU shortage that currently prevents OpenAI from releasing better models and meeting surging user demand.

• He called it "insane" that some startups with just "three people and an idea" receive massive valuations, predicting a significant market correction is on the horizon.

Bottom line: Altman's comments signal a coming market correction where hype gives way to utility. For professionals and developers, this means the companies with the deepest pockets and most robust infrastructure are poised to define the industry's next chapter.

Start learning AI in 2025

Keeping up with AI is hard – we get it!

That’s why over 1M professionals read Superhuman AI to stay ahead.

• Get daily AI news, tools, and tutorials

• Learn new AI skills you can use at work in 3 mins a day

• Become 10X more productive

Sign up and start mastering AI

AI Tools of the Day

1. 🚂 TheFluxTrain - Train your own custom AI image models with just a handful of photos to generate perfectly consistent characters and styles.

2. 🛡️ Sensity AI - Defend against digital deception with a powerful deepfake detection platform that verifies media to secure online identity and combat fraud.

3. 🎮 Jabali.ai - Design and build entire interactive games using a purpose-built AI engine that turns creative vision into reality without writing code.

4. ✍️ Haystack - Reimagine your coding workflow with a canvas-based IDE that visualizes function calls and simplifies complex project navigation.

5. 🔌 Eden AI - Access the best AI models from any provider through a single, unified API that simplifies development and orchestrates complex workflows.

6. 🎨 Albus - Unleash your creativity on an infinite canvas that integrates multiple AI tools for mind mapping, image generation, and document analysis.

7. 🔗 Contextual AI - Build production-ready RAG agents that securely connect to and reason over your company's complex internal data.

8. 🧠 Perplexity Comet - Experience the future of the internet with an AI-native browser that proactively automates tasks like scheduling and shopping comparisons.

Explore the Best AI Tools Directory to find tools that will 10x your output 📈

The Vanishing First Step

The Recap: As AI automates junior-level work, entry-level corporate jobs are disappearing, making it significantly harder for new graduates to start their careers in fields like tech. The first rung on the corporate ladder seems to be vanishing for many.

Unpacked:

• Hiring for new graduates at the 15 largest tech companies has fallen by over 50% since 2019, dropping from 15% of new hires to just 7% according to a SignalFire report.

• Artificial intelligence was explicitly cited as the reason for over 10,000 job cuts in the U.S. in the first seven months of 2025 alone, per new data.

• The challenge extends to the internship pipeline, with full-time offer rates for interns dropping to their lowest point in more than five years reports the National Association of Colleges and Employers.

Bottom line: Companies are embracing AI to cut costs and boost efficiency on junior-level tasks in the short term. This strategy, however, risks creating a significant skills gap by eroding the pipeline for future senior talent and experts.

AI Training

How to Connect WhatsApp to n8n (AI Workflow Tutorial)

The Recap: In this video, I'm going to show you the easiest way to set up a WhatsApp connection inside n8n to power your AI agents and automations!

P.S. We also launched a free AI Automation Community for those looking to build and sell AI Automations — Come join us!

AI's New Polyglot

The Recap: NVIDIA is boosting multilingual speech AI by releasing Granary, a massive open-source audio dataset, along with new models that support 25 European languages. This release helps developers build more accurate and inclusive voice applications.

Unpacked:

• Granary tackles data scarcity with one-million-hour of audio, providing a critical resource for languages underrepresented in AI, like Croatian and Maltese.

• The dataset was created using an open-source pipeline that automatically processes raw audio, helping developers skip resource-intensive manual annotation.

• NVIDIA also released two new models trained on Granary: Canary for high-quality transcription and translation, and Parakeet for fast, real-time speech recognition tasks.

Bottom line: This initiative lowers the barrier for building speech AI that supports a wider variety of global users. By providing both the data and the tools, NVIDIA enables the developer community to create more linguistically diverse applications.

Where AI Experts Share Their Best Work

Join our Free AI Automation Community

Join our FREE community AI Automation Mastery — where entrepreneurs, AI builders, and AI agency owners share templates, solve problems together, and learn from each other's wins (and mistakes).

What makes our community different:

• Real peer support from people building actual AI businesses

• Complete access to download our automation library of battle-tested n8n templates

• Collaborate and problem-solve with AI experts when you get stuck

Dive into our course materials, collaborate with experienced builders, and turn automation challenges into shared wins. Join here (completely free).

Google's Imaging Speedrun

The Recap: Google just made its advanced Imagen 4 text-to-image family generally available, introducing a new "Imagen 4 Fast" model built for rapid, high-volume image creation at a low cost.

Unpacked:

• The new Imagen 4 Fast model is designed for speed and high-volume tasks, costing just $0.02 per image.

• The family also includes the high-quality Imagen 4 and the top-tier Imagen 4 Ultra, with both now supporting detailed image generation up to 2K resolution.

• Developers can access the models through the Gemini API or test them in Google AI Studio, with all images watermarked using SynthID for responsible AI practices.

Bottom line: This move gives developers a flexible toolkit, letting them choose the right balance of speed, quality, and cost for their specific needs. It makes integrating high-volume, low-latency image generation into applications more accessible than ever.

The Shortlist

Anthropic proposed a transparency framework for frontier AI development, applying only to the largest developers to ensure public safety and accountability without stifling innovation.

Google announced its Imagen 4 family of text-to-image models is now generally available, introducing the new high-speed Imagen 4 Fast model and support for 2K resolution outputs.

New research revealed how popular generative AI tools create biased and racist caricatures of Australians, depicting a "typical Aboriginal Australian's house" as an outback hut while showing a suburban home for a generic "Australian's house."

Anthropic enabled its Claude Opus 4 models to end conversations as a last resort in rare cases of persistent user abuse, an experimental feature stemming from its AI welfare research.

Signing off,

David, Lucas, Mitchell — The Recap editorial team