Anthropic’s IPO is not simply another blockbuster technology offering. It may mark a fundamental turning point in capitalism itself: the moment intelligence begins its transformation from a scarce human capability into an industrial-scale, commercially produced resource. Just as electricity transformed virtually every industry in the 20th century, artificial intelligence is emerging as a foundational input to economic activity, one that can be manufactured, distributed, embedded in products, and deployed at extraordinary scale. Anthropic’s explosive growth is an early signal of this transition, but the real story is the vast economic ecosystem forming around it: semiconductors, data centers, electricity generation, enterprise software, autonomous agents, scientific discovery, and entirely new business models. The IPO offers investors a rare glimpse into the infrastructure being assembled to power what may become an entirely new economic order.

Yet there is a profound paradox at the heart of this transformation. We are investing unprecedented amounts of capital to manufacture a form of intelligence whose ultimate capabilities, economic value, and consequences we cannot reliably predict. Anthropic’s extensive risk disclosures reveal just how much remains unresolved, from the economics of massive compute commitments to the governance and safety of increasingly autonomous systems. This is what makes the IPO historically significant: it is simultaneously a bet on extraordinary productivity and a public acknowledgment of extraordinary uncertainty. The question is no longer simply whether AI will become a major industry. It is who will control the infrastructure of intelligence, who will capture the value it creates, and whether our institutions can adapt as quickly as the technology itself.

Share

Subscribe now

Leave a comment

Share

Subscribe now

Leave a comment

What happens when the technology advances faster than our ability to comprehend it?

Artificial intelligence is advancing at an extraordinary pace. But our understanding of it, our willingness to embrace it, and our ability to put it to productive use are moving at very different speeds.

And those differences may be just as consequential as the technology itself.

Over the past several months, I’ve been exploring a fundamental question: What happens when technological capabilities advance faster than individuals, organizations, markets, and institutions can understand, adopt, and adapt to them?

That exploration has evolved into The AI Gap Model, a working framework identifying 11 interconnected gaps across four major systems: human and organizational, economic, capital markets, and institutional.

At its foundation are three S-curves representing AI capability, human awareness, and real-world implementation. Together, they illustrate a central premise: The AI revolution isn’t unfolding at one speed. It’s unfolding at many speeds simultaneously.

These gaps create friction, uncertainty, risks, and extraordinary opportunities. They also help explain why perceptions of AI’s progress can differ so dramatically from its actual capabilities.

The model is a work in progress, not a finished theory. My intention is to share it openly, invite criticism and collaboration, and refine it as the transformation unfolds.

Coming soon to Deep Gap Journal: A deeper exploration of the AI Gap Model, its implications for society and capital markets, and what it might reveal about the road ahead.

This is a preview of a more substanitive paper that will be published on this Substack channel soon…

#ArtificialIntelligence #AIGapModel #TechnologicalDisruption #Innovation #DeepGapJournal

Share

Subscribe now

Leave a comment

For decades, using a computer meant learning its language: commands, menus, icons, applications, and increasingly complex interfaces. The graphical user interface, pioneered at research laboratories and popularized by Apple and Microsoft in the 1980s, revolutionized computing by replacing typed commands with windows, icons, and a mouse. Smartphones later made interaction even more intuitive through touch. But ChatGPT introduced another fundamental shift: instead of requiring people to learn how to communicate with computers, computers became remarkably capable of understanding how people naturally communicate. For the first time, millions of ordinary users could engage in open-ended conversations with a computer, asking questions, expressing ideas, refining instructions, and exploring unfamiliar subjects in everyday language.

This distinction is especially important for anyone who remains skeptical or intimidated by artificial intelligence. Before considering its extraordinary capabilities in research, reasoning, writing, analysis, and problem-solving, consider the simplicity of the interface itself. You don’t need to understand programming, master complicated software, or even know precisely how to formulate your question. You can simply begin a conversation. And increasingly, you can speak rather than type. The computer can ask clarifying questions, interpret your intentions, and help you discover possibilities you might not have known existed. This represents far more than another software innovation. It is a fundamental change in the relationship between humans and computers, one that has the potential to make increasingly sophisticated technology accessible to virtually everyone.

Three historical nuances worth getting right…

First, Steve Jobs and Bill Gates were instrumental in popularizing graphical user interfaces, but they didn’t invent them. Much of the pioneering work happened at Xerox PARC in the 1970s. Apple’s Lisa (1983) and Macintosh (1984), followed by Microsoft Windows (1985), helped bring graphical interfaces to mainstream computing.

Second, ChatGPT’s original November 2022 release was a text-based conversational interface. You still needed a keyboard. Natural voice conversations arrived later. The original breakthrough wasn’t eliminating typing; it was eliminating much of the need to learn specialized computer commands.

Third, natural-language interfaces existed before ChatGPT, including Siri and Alexa. What changed was the flexibility and breadth of the interaction. Instead of issuing a limited set of predefined commands, people could engage in open-ended conversations.

Share

Subscribe now

Leave a comment

GPT-6 Intelligent UI release

This month, OpenAI released ChatGPT 6. The GPT-6 Intelligent UI release represents a different kind of milestone.

Previously, software developers designed the interface and AI supplied the content. Now, ChatGPT can assemble an interface suited to the question itself.

The World’s Most Beautiful Libraries exploration is a modest example. Instead of receiving six paragraphs and six photographs, you received a navigable experience with buttons, images, facts, and a selection control.

The larger implication is that software may increasingly adapt to the user rather than requiring the user to adapt to software. That’s the direction OpenAI explicitly describes in its GPT-6 announcement.

And there’s an intriguing distinction for AI adoption, a topic I will be spending a lot of time exploring: improvements in model intelligence and improvements in usability can reinforce one another. A model doesn’t necessarily need a dramatic increase in reasoning ability to become dramatically more useful to a much larger population.

That could be one of the most consequential developments in this entire timeline.

The Evolution of ChatGPT

November 2022 – October 2026 | Models, Capabilities, and the Rise of Interactive AI

White paper | October 18, 2026

Executive perspective

ChatGPT evolved from a text-based conversational novelty into a multimodal system capable of reasoning, researching, writing software, analyzing images, and working with tools.

Two histories run in parallel: the GPT models (the underlying intelligence) and the ChatGPT capabilities layered around them, including browsing, voice, memory, image generation, data analysis, and autonomous research. A new capability does not always require a new model.

1. The major model releases

GPT-3.5 | November 2022

The conversational breakthrough

ChatGPT launched as a free research preview. It could converse naturally, explain concepts, write essays, draft emails, summarize text, and generate code. Its accessibility introduced generative AI to a mass audience.

Signature advance: Natural-language conversation

GPT-4 | March 2023

A major leap in intelligence

Much stronger complex reasoning, writing, coding, and instruction following. GPT-4 was designed to accept image inputs as well, although initial ChatGPT access was text-based.

Signature advance: More sophisticated reasoning

GPT-4 Turbo | November 2023

More context, more efficiency

A much larger 128,000-token context window, improved instruction following, more recent knowledge, and lower API costs. This made lengthy documents and more complex workflows practical.

Signature advance: Long-document comprehension

GPT-4o | May 2024

AI gains eyes, ears, and a voice

The omni model unified text, image, and audio capabilities, enabling much faster and more natural spoken interactions, improved visual understanding, and real-time multilingual communication.

Signature advance: Multimodal interaction

OpenAI o1 | September 2024

The reasoning revolution

Introduced an approach in which the model spends additional computation reasoning before responding. Particularly useful for difficult mathematics, scientific problems, programming, and analytical challenges.

Signature advance: Deliberate reasoning

OpenAI o3 / o4-mini | April 2025

Reasoning meets tool use

Models could combine deep reasoning with tools, including web searches, Python analysis, file interpretation, and image understanding. This made multi-step problem solving substantially more powerful.

Signature advance: Tool-assisted reasoning

GPT-5 | August 2025

Unified intelligence

Brought the strengths of conversational models and reasoning models together, automatically applying more thinking when appropriate. Stronger coding, analysis, and professional problem solving.

Signature advance: Adaptive intelligence

GPT-5.1 | November 2025

More efficient thinking

Improved adaptive reasoning, coding, responsiveness, and tool use. The model could devote less computation to simple questions and more to difficult tasks.

Signature advance: Smarter allocation of effort

GPT-5.2 | December 2025

Professional knowledge work

Improved spreadsheets, presentations, long-context comprehension, image interpretation, software engineering, and complex projects requiring multiple tools.

Signature advance: Professional deliverables

GPT-5.4 | March 2026

Computer-operating agents

Significant advances in coding, research, tool coordination, and computer use. Introduced native computer-use capabilities in developer environments and support for up to one million tokens of context in the API.

Signature advance: Software and computer interaction

GPT-5.5 | April 2026

Greater agentic independence

Improved the ability to take an ambiguous, multi-part task, plan it, operate tools and software, verify results, and carry work through with less supervision.

Signature advance: Extended autonomous workflows

GPT-6 | 2026 (month not independently verified)

Current flagship generation

The current GPT-6 generation emphasizes stronger reasoning, integrated tool use, interactive responses, and carrying complex work through to useful outputs. The precise initial public release month was not independently verified for this document.

Signature advance: Integrated reasoning and action

2. The capabilities that changed ChatGPT

Release

Major capability

Why it mattered

March 2023

Plugins and browsing experiments

ChatGPT began connecting with outside information and services.

July 2023

Code Interpreter

Upload data, execute Python, create charts, and analyze spreadsheets.

October 2023

DALL·E 3 integration

Generate images directly from conversational descriptions.

February 2024

Memory preview

Personalization could carry information between conversations.

May 2024

GPT-4o voice and vision

More natural voice conversations and multimodal understanding.

January 2025

Operator

An AI agent could operate a browser using clicks and typing.

February 2025

Deep Research

Autonomous, multi-source research with citations and synthesized reports.

July 2025

ChatGPT agent

Broader combination of research and browser-based actions.

October 2026

Intelligent UI

Interactive charts, controls, diagrams, and tailored visual interfaces directly in conversations.

3. The bigger picture: four eras of ChatGPT

2022–2023: Conversation

The breakthrough was making powerful language models accessible through ordinary conversation.

Transition: From answering questions to understanding the world.

2024: Multimodal intelligence

AI gained stronger abilities to see, hear, speak, and reason through difficult problems.

Transition: From understanding to performing work.

2025: Reasoning and agents

ChatGPT increasingly became capable of researching, planning, using tools, and completing multi-step tasks.

Transition: From doing work to shaping the interface.

2026: Interactive and agentic AI

Models became better at operating software and coordinating complex work, while Intelligent UI made responses directly interactive.

Transition: From static answers to dynamic, adaptive experiences.

4. Why October 2026 is particularly interesting

The GPT-6 Intelligent UI experience represents a different kind of milestone. Previously, software developers designed the interface and AI supplied the content. Now, ChatGPT can assemble an interface suited to the question itself.

The library exploration offers a modest example. Instead of receiving six paragraphs and six photographs, the user received a navigable experience with buttons, images, facts, and a selection control.

The larger implication is that software may increasingly adapt to the user rather than requiring the user to adapt to software.

Improvements in model intelligence and improvements in usability can reinforce one another. A model does not necessarily need a dramatic increase in reasoning ability to become dramatically more useful to a much larger population. This may be one of the most consequential developments in the timeline.

5. Implications for adoption and interface design

The story is not simply a sequence of smarter models. It is a shift in the unit of value: from answers, to analyses, to deliverables, to workflows, and ultimately to personalized interactive experiences. Lowering the interface barrier can expand adoption independently of raw model benchmarks.

Source and verification note

This paper preserves and expands the preceding conversation. Release dates and capability descriptions are a historical overview, not a freshly audited release ledger. In particular, the prior conversation contained conflicting claims about the first public release month of GPT-6 (unverified versus September 2026). The exact GPT-6 release month and the October 2026 Intelligent UI launch claim should be checked against official release notes before external publication.

Primary sources for verification: https://openai.com/news/ ; https://help.openai.com/en/articles/6825453-chatgpt-release-notes ; https://openai.com/index/gpt-4/ ; https://openai.com/index/hello-gpt-4o/ ; https://openai.com/index/introducing-openai-o1-preview/ ; https://openai.com/index/introducing-deep-research/

Share

Subscribe now

Leave a comment

I have spent most of my life exploring. Sometimes that has meant mountains, trails, cities and unfamiliar places. More often, it has meant ideas: business, markets, technology, organizations, human behavior, customer experience, education, communities and, increasingly, artificial intelligence. I’ve always been curious about how things work, why they work, why they don’t, and what might be waiting just around the corner. All of that will live here on the Deep Gap Journal.

I will use this new Substack publication to share my thoughts, observations, conversations, essays, papers, infographics, photographs and other things I discover along the way. Increasingly, some of those explorations begin as conversations with AI. Those conversations have become an extraordinary way for me to question assumptions, connect ideas, test conclusions and venture into subjects I might never have explored otherwise. When something useful comes out of that process, I’d like to share it.

The Deep Gap Journal will also provide a home for my previous works. Over the years I have created a substantial body of work across several different chapters of my life: publications, articles, presentations, business programs, models, videos and other projects. Some of it is decades old. Some is much more recent. Rather than leave those pieces scattered across old websites, publications, computers and digital filing cabinets, I am gradually bringing the work that still has value together here. Some will remain as it was originally created. Other pieces may be revisited, updated or completely reimagined with the benefit of hindsight and today’s remarkable tools.

That is part of the meaning behind An Explorer’s Journey. The old work is not separate from what I’m doing today. It’s an earlier trail. It shows some of the places I’ve been, questions I’ve asked, ideas I’ve pursued and lessons I’ve learned. The new work simply picks up the trail from here.

The name Deep Gap comes from the Deep Gap Trail, which has become one of my go-to hikes here in the Asheville area. I’ll be sharing more thoughts about this magnificent trail soon.

I also like what the words suggest. To go deep is to look beneath the obvious. A gap is a passage between places, and sometimes between ideas. I’ve found that some of the most interesting discoveries happen when you wander into those spaces and start asking questions.

And questions may ultimately be what this journal is about. I have never been particularly good at accepting that’s just the way it is as an answer. I like looking at familiar things from unfamiliar angles, connecting dots that don’t necessarily appear connected, and asking what might be coming next.

At this point in my life, I have covered fairly long trail but I am much more interested in the trail that lies ahead.

The Deep Gap Journal will be both a record and an ongoing exploration: a place to preserve work worth keeping, revisit ideas worth reconsidering, and share new observations as they emerge.

I don’t know exactly where it will lead but I promise it will be interesting.

Thanks, again, for stopping by. New content should begin to appear in late September 2026.

Steve Saenz
September 15, 2026
Asheville, NC