Hello fellow humans! Today, we share an amazing AI bubble analysis from The Neuron, followed by valuable insights on what skills we need as AI becomes integrated into our lives, and an outstanding examination of how educators can think about AI in the classroom and focus on developing cognitive skills that will serve students with and without AI.

Please check out the full reports.

The Neuron has a great in-depth analysis of the bull versus bear case for whether or not we're in an AI bubble. I strongly recommend you read the whole thing at The Neuron, but here’s a quick summary.

The bull case comes from Marc Andreesen and other VCs betting big on AI.

It’s worth noting that the strongest bull cases come from people with a vested interest in AI succeeding. They imagine a per-token pricing model that costs around $100k/year for a full-time software developer, which is a lot more than AI costs now, but still a lot less than a full-time human developer.

UK analyst Julien Garran calls this “the biggest and most dangerous bubble we’ve ever seen.”

This assumes that the pricing model doesn’t change much and that customers will not sign up for pricing structures that actually pay for the compute costs.

The Neuron agrees with the bulls that technology is revolutionary and is here to stay in one form or another. But it’s also clear that AI is not generating enough revenue and is running out of VC runway. The frontier firms will need a new business model. The question is how the frontier firms will price access for founders, investors, and professionals. The web survived the dot-com bust, and AI will survive this bust. The question is: what will survive the bust?

As long as LLMs remain the dominant AI model type, AI will not replace humans to any meaningful degree. Further, as the cost of energy to operate data centers will continue to increase in the US, and as VC funding dries up, the cost of buying AI inference at scale from frontier models will become a hurdle for AI startups and end-user customers.

Over the next 12-48 months, chipmakers will start delivering cheaper, faster, and more energy-efficient chips for in-home or in-office AI inference. This local AI compute capacity may be an opportunity for software builders to offer smaller, more specialized AI models that can use local AI compute instead of relying on the major model operators. Those companies that can genuinely benefit from AI will have local solutions (think open-source or open-weight versions of major models) that they can manage with context engineering.

There will be a market for a works-out-of-the-box solution from Microsoft and Apple that functions as an assistant for various professional functions, but will still rely heavily on human skill to produce useable work with LLMs.

When we have a successor to the LLM, that may be the inflection point that shows us more of what the post-AI-bubble world will look like.

The consulting firm Hemsley Fraser published a paper describing key human capabilities and skills needed in the workplace, with or without AI. It describes four capability categories with 3 key skills in each capability.

At a glance, these seem like “soft skills” that might be downplayed in the job market, but employers spend as much time validating these interpersonal skills during the interview process as the technical skills.

There's growing concern about maintaining essential human skills while leveraging AI capabilities. Organizations are focusing on developing "uniquely human" skills like creativity, interpersonal abilities, and critical thinking while preventing cognitive atrophy from over-dependence on AI systems. A new report from the Institute for Management Development found:

The report identifies six functions that AI tools need to support in human learning and cognition to be allowed in any education program:

The report gets more detailed with 10 things AI support should do and 10 things it should not do. According to the report, the challenge is to find the right balance and suggests a 75/25 split between traditional instruction techniques and AI integration.

The report also provides practical strategies for how to incentivize engagement with the relevant learning materials and how to align your assessments with future-relevant skills. This report is as relevant to high school and college educators as it is to executive leadership trainers.