I will get right to the point.

This is a profiler snapshot of Google's gemini website when pasting in a large chunk of text.

What you are seeing is the UI freezing for a solid 10s. If I paste a larger chunk of text, the browser will crash.

This is not an obscene amount of text, a few thousand lines. For context, LinkedIn's search input is instant with the same text as is most of the web.

Google is spending billions on AI paying people million $ comp plans and they cannot seem to get an html input element right.

What is going on?

Is this a result of those pesky zoomies vibe coding? Is it the crappy web devs who think they are engineers? Or just laziness?

After, looks at watch, 3 minutes of debugging, I found the issue.

Google's gemini web app uses rich textarea an obscure react component which will crash your browser if you paste more than a thousand lines of text.

Excuse me, they do not use that but a library called quilljs which is much more powerful and complex but it too also crashes if you paste too much text into it. :/

I know what you are thinking. Web devs just don't know how to make quality software. Perhaps. But I don't blame the author of the component but Google.

it is the web devs fault

I am something of a web dev (moonlighting as an engineer) and we get a bad rep, often for good reason. But the author of the rich textarea component made OSS for free. And Google is using it to make what is ostensibly a billion $ product without paying them.

Yes yes, the real product is gemini and not a janky little web app but my point stands.

Is Google seriously so strapped for cash they cannot make an input field that handles > 1k lines of text? Can they not spare a few million? Gosh, I bet the OSS peeps would take the time to improve their code for a cup of coffee if offered.

no really, it is the web devs fault

I get it. It is fun to poke at web devs with 5mb js bundles and our overly complicated SPA's with silly animations that eat battery life but come on now, are we really to blame here?

And no it isn't the vibe coders who get excited about an LLM which decides to use rich-textarea or quilljs or whatever.

"Engineers" at big tech companies can be so gosh darn lazy. I know I am.

I could certainly justify the decision of using random lib rather than spending the extra week or two of effort building something from scratch. I'd even use data to back the decision. "only 2% of queries use more than 300 tokens so do we really need inputs that support 150k tokens?" Gosh, if I added a few charts and a slide deck, I could even get a promotion.

Look ma! I saved us 100k in eng hours by using FREE oss instead of doing my job...or maybe that is my job, to just use oss software...

can we have nice things?

Yes, yes we can but you got to want it. Google, doesn't want it. They don't want nice things. It is why they have a graveyard of amazing failed products. It is because deep down, the culture at Google, (and I'd argue most of big tech) is about...not caring. It is about doing the absolute minimum possible so one can buy another rental property, supreme brick, stonks, or whatever people spend money on.

Crypto? It has to be crypto. How else is that thing still around.

be honest

I criticize myself here as much as I do Google or anyone else. I am part of the problem. I am part of the culture. Am I going to make a PR to rich-textarea to fix it to be more performant? No, no I am not.

Because I don't want nice things, or at least, I don't want a nice gemini web app. I'd rather write a blog post complaining about it.

Perhaps dear reader you do want it. I certainly hope so, otherwise, gemini is gonna keep crashing on me.

Do note, I am forced to use gemini by my company and as an LLM, it isn't bad, but as usual, Google's execution is...well...googly.

Until next time.

where to find me?