iiinova — where your digital self keeps evolving

Like you use fable 5 for everything? I'm finding a bunch of easier UI tasks that sonnet handles just fine - I'll spawn off an autonomous agent which commits it's work on a branch and if I don't like the outcome I just delete and try again with a better prompt/model.

410

Yes indeed, and it's fun to have to learn to really optimize again too. Reminds me of coding as a kid when I couldn't afford to upgrade my 640kB RAM 😛. Though I really hate the tradeoff of choosing a worse model which then creates bugs which wastes my time (vs. cheaper meaning better engineered).

Yeah but just pro. My hobby coding is intermittent and I do still try to use Gemini sometimes too. But this month I definitely would have been better off on max. On the plus side, waiting for my 5h token limit to refresh is forcing me to take a break and eat lunch 😛

Rick Byers @rbyers.net · 10/07/2026

Oh oh, I seem to have become addicted to Claude tokens! Probably late to the party, but I keep upping my spend limit. I'm at $400 so far this week 😬. This is because I thought I could build my own version of something that cost $2800 to buy, we'll see if I end up saving money or not 😊.

260

Chromium costs hundreds of millions of dollars a year just to keep working well. For better or worse, none of us really believed there was a credible way to make a dent in that via charity.

020

Rick Byers @rbyers.net · 10/07/2026

But there are lots of ways for individuals to contribute their time and expertise to chromium, and there are lots of other worthy projects like Servo and Ladybird where small donations can go a lot further.

220

Rick Byers @rbyers.net · 10/07/2026

Yeah we discussed it some. Personally I don't think it's fair (or good optics) to ask individuals to donate money to Chromium without getting anything directly out of it. Maybe if we could find a scalable way to pay for getting pet peeve bugs fixed or something.

Rick Byers @rbyers.net · 02/07/2026

This is the #1 thing ChromeOS got right (and Android copied) - reserve space for 2 read-only boot partitions so OS update is a reliable and trivial active partition swap.

Rick Byers @rbyers.net · 02/07/2026

Yeah. In this case we had "automatically update" enabled but there was insufficient disk space, so it took some effort to actually get an OS update installed.

Rick Byers @rbyers.net · 01/07/2026

I am a negligent husband! I just discovered my wife's MacBook was running MacOS 10 and so stuck on Chrome 128 and missing 2 years of security patches! If even I fail to keep my family safe online we might be doing something wrong 😕.

270

Rick Byers @rbyers.net · 23/06/2026

Thanks, yes exactly. But there is a legitimate argument that just making that "cheaper" has a potential privacy cost: github.com/mozilla/stan.... Personally I believe it's certainly a cost worth paying, because it's also easy for us to target interventions if necessary.

github.com

CPU Performance API · Issue #1364 · mozilla/standards-positions

Specification title CPU Performance API Specification or proposal URL (if available) https://wicg.github.io/cpu-performance/ Explainer URL (if available) https://github.com/WICG/cpu-performance Pro...

010

Rick Byers @rbyers.net · 20/06/2026

Vibe coded a demo of the new navigator.cpuPerformance API in Chromium: rbyers.github.io/cpu-performa.... It's a small thing (returns just 0-4), but for some powerful web apps I think it'll be super useful. Details: github.com/WICG/cpu-per...

rbyers.github.io

CPU Performance API — Reference Implementation Playground

151

Rick Byers @rbyers.net · 19/06/2026

CSS Quake (cssquake.com) - no canvas just HTML/CSS using a cool library called PolyCSS (polycss.com). 😀 Playing through the first level brought back a lot of memories. I think I was one of the first to reverse engineer the PAK file format and build an editor days after the game came out.

cssquake.com

cssQuake - Play Quake in Your Browser

Play Quake in your browser with no install. cssQuake renders the game as inspectable HTML and CSS powered by PolyCSS.

1101

Rick Byers @rbyers.net · 16/06/2026

Given how Chromium and WebKit are always vying for the top spot in Speedometer on MacOS, it's really striking how big the gap is on iOS! And we haven't even really tried to optimize performance for that platform yet! IMHO this is what you should expect to see when there's a lack of competition!

062

It's a good question. I don't know the team's plans but I'd guess we'd want to expose OS APIs where possible and high quality. For now, given performance constraints, the prompt API is launching on desktop only where AFAIK good native OS models do not yet exist?

010

Android shipped the Google AI Edge SDK in Oct 24, and iOS got the Foundation Models Framework in Sept 25. So this is not at all about the web being first, but once again trying to reduce the risk of loss of relevance by not being too far behind.

321

Yeah. Also there probably is some small entropy leakage about the user's available disk space. But I agree, this all seems less important than the other concerns you raised.

120

A great analysis! The punch lines: "The trend is clear: for every 100 milliseconds slower a store loads, conversion tends to be about 3.5% lower." "For every 32 milliseconds slower a store responds to interactions, conversion tends to drop by about 1.5%."

Chart with caption: Stores with faster LCP consistently show higher conversion rates. Even when controlling for other performance factors, the relationship remains strong.

041

Great data, thanks for sharing! Is there a reason you don't share the CLS chart along with LCP and INP? IMHO negative results are just as important as positive ones (if not moreso) in helping us learn how things really work...

100

I think if you want to try to guess if you're behind the great firewall of China, there's a ton of things to try ahead of attempting a gigantic model download. Eg. I assume Google fonts are also blocked. I'm not aware of any reason why the model would differ for different people.

120

Mozilla's Vision of the Web

Thanks. In Mozilla's defence, their resources are quite limited so they have to be very selective in how they invest. And they do make some unique investments (eg. the Rust language), prioritized by their vision for the web: www.mozilla.org/en-US/about/....

mozilla.org

040

Rick Byers @rbyers.net · 30/04/2026

Google services largely don't operate in China AFAIK, so no model download at all (if you can get even Chrome installed given no plat store). I'm proud that the company would rather be blocked as illegal than compromise on principles like striving for the most accurate information.

150

Rick Byers @rbyers.net · 30/04/2026

I.e. @jakearchibald.com I've already admitted you were right and I was wrong in one case (view transitions vs. portals / high-vs-low level web APIs). Will this be one where we even the score or will you double your money? 🙂

Rick Byers @rbyers.net · 30/04/2026

There's some legitimate concerns here. Let's try to be concrete about how we will identify the harm in practice so that years from now we can all learn from this difference of perspective. github.com/mozilla/stan...

github.com

Prompt API · Issue #1213 · mozilla/standards-positions

Specification title Prompt API Specification or proposal URL (if available) No response Explainer URL (if available) https://github.com/webmachinelearning/prompt-api/blob/main/README.md Proposal au...

140

Reposted by Rick Byers

Sam Goto @sgo.to · 30/04/2026

I think part of the phenomena here is that it is easy to predict things will fail: most things fail. So, stop energy is much cheaper to produce, because you would be right 99% of the time. Against-everything-for-nothing is a comfortable stance to take, but long term it is the 1% that move forward.

322

Thanks! Exactly, what do I need a 100MB native app running all the time for my mouse for when 10kB of JS is enough to get the job done?

010

Oh it works, just the UI looks bad on phones. I'm sure Gemini or Claude would have no trouble fixing it if I ask for a responsive design. And yes exactly, why constantly run an app using 100MB of RAM for something so simple?

000

Probably some way, though I assume the details are different. Claude is also pretty good at reverse engineering binary executables if you give it a disassembler and debugger!

000

I just thought to test the UI on mobile. I guess our jobs are safe for a little while longer 😂.

A web app with poor layout and excessive wrapping on a phone-sized screen.

140

Next step: vibe code a replacement for concur so I can file my BlinkOn trip expenses in 1/10th the time? 😜

000

Just bought a new Razer mouse. Rather than install the native app, I had Gemini/Claude write a web app with the Web HID API, using the Linux driver as a reference. Less than 60 min of work laterand I think I'm done: razerweb.rbyers.ca It's a new world... Software on-demand!

razerweb.rbyers.ca

Razer Web Configurator

7214

In this case it's a repeat call from one company in particular which previously scammed my family member out of $200, hoping to repeat their success. I'm frustrated that VISA has so far refused to issue a chargeback, even though I have photos like this documenting the obvious scam.

100

I figured a scam will be left either trying to convince the victim to use a different communication mechanism, will be forced to give the address of a legit company whose contact form won't reach them, or will risk Google Safe Browsing block and takedown action.

100

My best suggestion: "I'm busy right now but if you give me your website address, I will reach out through your contact information there". Don't accept a phone number or other form of contact. Any legitimate IT company will have a website with contact info that works.

Someone in my family asked what they can say to a tech support scammer calling them to confirm they are a scam, not a legitimate company. Thoughts?