Google (GOOGL) Feels Like A Missed Opportunity

May 24, 2026

I watched some of Google’s I/O sessions. Google GOOGL 0.00%↑ has cooked up a lot of good stuff, yet it is clear the company is building a walled garden - much like Apple AAPL 0.00%↑ did many years ago - which I find uninspiring. Google cares more about its walled garden than having a best-in-class frontier model. I believe that is a mistake.

Take Google Spark, which is Google’s answer to OpenClaw and Hermes Agent. The big difference is that Spark only connects to Google products whereas other agents (OpenClaw, Hermes, Claude, Codex, even our Kilby agent), connect to multiple 3rd party APIs and tools. Broad horizontal automation or deep vertical automation (our path), is the very point of building an agent. Google Spark is neither.

Spark will eventually have hundreds of 3rd party connections, but why launch it if it is sub-standard? Why not wait 6-12 months? I know the answer. Many people will be satisfied to simply link Gmail, Calendar and Google Drive, and why not get them hooked today? Personally I like to read my emails and would never set an agent to send me summaries and not read the originals as too much nuance is lost with summaries, which could mean lost revenue.

Many who love Google will use Spark just as many who are happy being captive in Apple’s Jurassic park, I mean “walled garden”.

Another element working in Spark’s favor is that Spark is ready out of the gate for the average person who is not comfortable running an agent on their local machine or setting up a virtual machine.

Spark is available on Google’s Ultra plan, which costs $100 month for 20 TB of storage and 20x more Gemini usage than the Google’s Pro plan.

Below is Google’s 3.5 Flash (flagship model) pricing. Cheaper than Claude and Codex in terms of token pricing, and a very capable model - althogh not so capable for writing code. However, if you care more about automating your email, calendar and getting max storage, Google’s Ultra Plan may be the right plan for you at home and at work.

Gemini is a sub-par coding model. If I were sitting in Sundar Pichai’s chair I would cancel Google I/Os until we had the best frontier model for coding. I would allow innovation in other product areas to suffer. I would do everything in my power to have the best frontier coding model in as short a timeframe as possible and to extend that lead over time. If that meant relentless pursuit of Dario Amodei and Anthropic’s world class research lab, which just scooped up Andrej Karpathy, then so be it. Dominate the frontier model space, and everything else becomes easy.

My sense is Google never sought to be best in class in the frontier model space, they just want to be good enough so that their customer base will opt in to using Gemini and Spark across Google Workspace. Very boring and a missed opportunity if you ask me.

The opportunity is still there, but Sundar is not the right guy to lead Google. If he wants to attract the best and brightest to Google, he’s got to project fire in the eyes and in the belly, even if doing so is against his personality. He’s got to publicly project a “win at all cost” attitude if he wants Google to win.

It is possible that the math may work in Google’s favor. The point may arrive where the cost of models and data centers is so great that Anthropic and OpenAI can’t survive on their own. Perhaps Google is patiently waiting to pounce. I do not agree with that approach either. I believe that if Pichai was clear about wanting to dominate the frontier model space, Google would have better models, a better research bench, and an even higher market cap. Perhaps GOOGL would be the world’s most valuable company. It all feels like a missed opportunity to me, especially with xAI essentially dropping out of the frontier model space by way of leasing chips to Anthropic.

At least Pichai is honest about Gemini’s laggard position in the interview below.