This is a submission for the Google Cloud NEXT Writing Challenge

Introduction

One announcement that caught my interest at the Google NEXT 2026 event was the announcement [Kim 2026] of the feature that:

“Auto browse with Gemini in Chrome...”

This fascinated and intrigued me, so I dug a little deeper with this feature and used Gemini in Chrome. I tested this auto-browse feature on Chrome and prompted Gemini to see what this new feature can do.

This essay is not one specific thing, but a pastiche or medley of many things. Specifically, something tried, reflection upon, opinion piece, and hopefully some insights and perspectives on where we are now, and what is potentially the future from some of the announcements and presentations at Google NEXT 2026.

Bona Fides Previous Work

However, before going into the details, observations, insights, perceptions, and my technical opinion, I will share my bona fides--my abilities, background, experience, and skills before going any further. I will answer the question (without using AI), “Whom is this person, William Fletcher Gilreath, and what does he know about anything in the AI/ML/LLM space?” Currently, with all the buzz about AI and ML, it seems everyone is an “expert” with AI and ML.

For this digression, I’m a senior engineer, both in software and machine learning, along with a computer scientist and mathematician. I have published theory papers and technical articles that are accessible on my home site [Gilre 2026].

In the 1990s at university, I worked with a professor doing AI and machine learning using neural networks, genetic algorithms, and simulated annealing. A topic of interest was swarm intelligence in distributed systems. One theory paper I wrote was a project to evolve pseudo-random number generators using a genetic algorithm.

I worked on several projects with this professor--such as evolving a combat schedule for units in sixty seconds, and taking a satellite image into a map. My previous job role was as a senior machine learning engineer at VMware, and I worked on Project Crest [Ambie 2021], using AI and ML to automatically detect issues with content accessibility.

For fun, I built a parallel super-search engine (https://get-star.org) [Gilre 2025] that does parallel in-tandem search for a query. One variation is EnneAiMl (https://get-star.org/enneaiml.html), which sends a prompt to query multiple AI sites with different models. I wrote, “PaLM shell” [Gilre 2024], which was published in Linux Magazine. The shell is a bash shell script that allows you to prompt the Gemini model in an interactive read-prompt-print-loop.

Thus, I am not a newbie or a “starlet” in the AI and ML space, but an experienced professional.

Playing with Chrome using AI

I double-checked that Chrome was up to date and that the AI option was present. From verification, auto browse was in Chrome. I then tried several tests, as an experiment to test the auto browse feature. I will not prolong the discussion by going into specific details, prompts, and the results (although I did copy and paste to a text file and saved to a link...), but I will share the insights from some of the tests I did.

• one-instruction set computer (OISC) - I did a search with Chrome on a technical speciality of mine, one-instruction set computers. Having written a computer architecture book on the topic (https://www.caamp.info), I am experienced and knowledgeable enough to assess the results of the prompt.

A web search returned websites, many of which I have visited, with OISC projects and comments. Using Gemini, I also found content, but it was a summary, description of OISC concepts, ideas, and links to references. One feature I did like was that I could further refine the results, and Gemini asked me if I wanted to do something more specific--prompting the user.

• essay outline and structure - I opted to do a “cheat” (well, in academia...) of a possible essay about LLMs. I prompted Gemini with the idea, asked for an outline, and suggested content. Gemini gave a very useful result of an outline, the topics to touch upon and discuss, and a summary of each part of the essay. This saved me the time and effort to do so. But Gemini gave me a head start, so I could focus on the content and flow of ideas and concepts.
  When I did a web search, I found related material, such as research papers, websites, and documents, so the search narrowed down the content I could utilize. But no outline, ideas for content, just references.

• source code - As a senior software engineer, I was curious to see if I searched for a data structure (one I created for a transpiler I am writing--indexed map in Java...) and what the result was from a web search. I found some existing implementations, but those implementations lacked some features or were generalized. I would have to implement around the existing data structure (one from Apache Commons) to achieve the data structure with the features and properties I wanted.

I then prompted Gemini with a description of the data structure. And then refined my prompt for the data structure. Eventually, Gemini gave me a Java class that used Java collections and implemented an indexed map. The ability to refine the results, either prompted by Gemini or by my own self from inspection and observation of the generated Java source code, helped refine the results. From the vibe generated source code, I could then focus on refining the generated Java class for some optimizations to improve efficiency or performance.

I did several other tests, but I will reserve those details for this discussion, perhaps another technical article, or research paper in the future. 😜

Programmer as Navigator

My efforts and tests with web search, and then using the auto browse or Gemini in Chrome, reminded me of a theory paper when I was studying databases in college. The theory paper was the 1973 Turing award lecture [Bachm 1973] by Bachmann.

In practice, Bachmann described searching a database by a user as a programmer. The programs are the database-related queries to search the data set. I will avoid a digression into the history and features of database systems.

The concept of programmer as navigator from the title seemed germane to using the auto browse feature, Gemini in Chrome. The programmer is the user, and instead of writing a program, the prompt searches...and the database is the Internet.

This is the software/database engineer's perception of the auto browse in Chrome with Gemini. But not everyone (much to my astonishment at times...) is a software engineer, machine learning engineer, or even technology savvy.

Yet this realization had me cogitating and pondering, and reflecting (more thinking while I was going through my daily routine...) as the auto browse reminded me of a concept, vaguely. I was in a webinar, and then suddenly the concept struck me--the concept of a “knowledge navigator” from the late 1980s.

The “Knowledge Navigator”

In his book [Scull 1987] “Odyssey from Pepsi to Apple”, the former Apple CEO John Sculley, near the end, writes about a concept he called the “knowledge navigator” which seems initially a present prediction of the World Wide Web in the early 1990s. Yet it was the concept of a user navigating knowledge.

Sculley [Scull 1987a] describes the knowledge navigator as:

“...a wonderful fantasy machine called the Knowledge Navigator, a discovered of worlds, a tool as galvanizing as the printing press. Individuals could use it to drive through libraries, museums, databases, or institutional archives. This tool wouldn't just take you to the doorstep of these great resources as sophisticated computers do now; it would invite you deep inside its secrets, interpreting and explaining--converting vast quantities of information into personalized and understandable knowledge."

Converting vast quantities of information into a personalized and understandable form is a natural task for an LLM. Or as Sculley [Scull 1987b] elaborates:

"Most important, the Knowledge Navigator will customize knowledge for you-it learns as you use it-to make navigating through information and ideas as interesting and understandable as possible."

The knowledge navigator is not just to access information, but more to make the user an explorer, as Sculley [Scull 1987c] explains:

"...the world will have a tool that will provide not mere information, but true knowledge, cheaply and efficiently, for all the ‘out’ and the ‘ins’ willing to use it. People will have the power to wander through centuries of knowledge as true explorers. We can skim the surface, occasionally diving in to incredible depths when we choose."

An interesting insight given [Scull 1987d] is about AI, and agents in the knowledge navigator:

"Artificial intelligence will play an important role in the Knowledge Navigator. Inside the soul of the computer will be intelligent software ‘agents’. Over time, they will become smart enough to learn that you like certain types of information presented in certain ways. The agent will learn along with you and work invisibly, turning information into useful knowledge for you."

The Current State of Things

The current state of the world is that the database is the Internet, the data is online, the user is navigating on their device, and the knowledge navigator is now approximately some forty years later.

Technology and knowledge have reached an inflection point in terms of information, with the search for information within the massive, dizzying amount of knowledge that is available on the Internet. But a question or inquiry is “Why now?” What was needed is...

1. A vast trove of information that is readily accessible on the Internet.
2. The AI intelligence, as an LLM to access the knowledge and information.
3. An app that is a web browser for a user to utilize to access information.

The confluence of these three needs, and now with Chrome + Gemini, has created the knowledge navigator.

Going forward into the future, the trend is that an increase in old knowledge is put on the Internet, and the content that users are creating and putting online. Thus, more knowledge, more information, more discovery. This will drive the need and necessity of the knowledge navigator to access that information in a simple, consistent way.

The union of Gemini as the AI, and Chrome as the app, as the web browser creates a knowledge navigator...and none too soon.

Conclusion

One strange aspect of ideas...is that they take time for technology to catch up to make it possible.

Consider that Thomas Edison did not invent the light bulb, but he did find a way to make it mainstream and widely available. That is the potential case here: Chrome + Gemini is making the "knowledge navigator" concept mainstream.

The "memex" written about by Vannevar Bush in 1945 did not reach some fruition until 1990 with Tim Berners-Lee and the advent of the World Wide Web. So from 1945 until 1990, approximately 50 years. The "knowledge navigator" posited by Sculley [Scull 1987] in his book in 1987 until now, in 2026, approximately 40-years.

Yet with Gemini + Chrome combination leads to the concept of the "knowledge navigator" only in the browser, using an LLM, and the massive database of the Internet. Ironically, the same problem Bush observed [Bush 1945] from scientific progress is that...a "growing mountain of research" that was becoming impossible for any single person to grasp, remember, or connect, leading to knowledge being lost.

The irony is the same problem from 1945 to the web browser and web in 1990, and now in 2026, the same problem. Yet the solution is the "knowledge navigator," and the combination of Chrome + Gemini in the web browser perhaps is the next step of a new era.

This is the fun, as time will tell, and whether it is a new era of knowledge navigation, or an ideological cul-de-sac. We will see as history and technology unfold. Excelsior!

References

[Ambie 2021] Ambiel, Suzanne. “Do I Make Myself Clear” https://blogs.vmware.com/opensource/2021/02/25/do-i-make-myself-clear, February 25, 2021, Accessed April 25, 2026.

[Bachm 1973] Bachmann, Charles. “The Programmer as Navigator,” Communications of the ACM, Volume 16, Number 11, November 1973. https://people.csail.mit.edu/tdanford/6830papers/bachman-programmer-as-navigator.pdf, Accessed April 25, 2026.

[Bush 1945] Bush, Vannevar. "As We May Think," The Atlantic Monthly, July 1945, pp. 112-124. https://www.math.cmu.edu/~af1p/Teaching/INFONET/Papers/AsWeMayThink/bush.html, Accessed April 25, 2026.

[Gilre 2024] Gilreath, William F. "Smart Shell," Linux Magazine, Issue 282, May 2024. https://www.linux-magazine.com/Issues/2024/282/Bash-PaLM-Shell, Accessed April 25, 2026.

[Gilre 2025] Gilreath, William. "Get Star: The Problem of Finding Information across Multiple Web Site," https://get-star.org/get_star_whitepaper.pdf, Accessed April 25, 2026.

[Gilre 2026] Gilreath, William. “William Fletcher Gilreath Home Site,” https://wilgil.me, Accessed April 25, 2026.

[Kim 2026] Yulie Kwon Kim, "10 more announcements for Workspace at Google Cloud Next 2026," https://workspace.google.com/blog/product-announcements/10-more-announcements-workspace-at-next-2026, April 22, 2026. Accessed April 25, 2026.

[Scull 1987] Sculley, John, and Byrne, John. Odyssey Pepsi to Apple, HarperCollins Publishers, New York, New York, United States of America, 1987, pp. 540-541. https://archive.org/details/odysseypepsitoap0000scul_h6k4, Accessed April 25, 2026.

[Scull 1987a] ibid, pp. 540-541.

[Scull 1987b] ibid, p. 541.

[Scull 1987c] ibid, p. 545.

[Scull 1987d] ibid, pp. 547-548.

[Villu 2021] Villuri, Suhash. “Brown Wooden Handle on White and Brown Textile”, February 2, 2021, https://unsplash.com/photos/brown-wooden-handle-on-white-and-brown-textile-DYWHwLZyJ64, Accessed April 25, 2026.