One of the ways on napkin math certainly is more expensive.

It has copied all the features it just does them all a bit worse.

Email filtering, ticket triage/categorization/etc.

It would be awesome to be able to create mail filters using natural-ish language instead of grammar

If this were true you could do a talk now and measure it and do it every week and see if it was still true or not and be the first to catch them in the act!

You can get ds flash providers that do 1000+ toks/sec - it’s unreal fast.

Distilled, weaker versions though I think?

Gemini Flash was my daily driver until 1017 dropped. Not that 1017 is better, they both are dumb in their own ways, it’s just available at far higher tok/sec.

Gemini models also historically are the best writers. 3.1 was the first model I saw that made actually funny jokes regularly.

Anthropic has had a laser focus on Enterprise features whereas Chatgpt is much more consumer focus.

Is anyone actually using those features? I feel like this still very much just Codex vs Claude code, and if codex has the better model everyone will switch in a week.

All the other products feels like a (desperate?) attempt to create moat and friction for switching models.

I feel like all these versus arguments are absolutely opinion and preference, but really deep down do hint at a modification that will be hard to ignore as an investor.

If AMD and Nvidia could make the exact same thing the exact same rates, why would the stock price of one be higher than the other?

Yes, they are. You don't just give 1000 employees a subscription and tell them they can have as many tokens as they want.

Apple is very consumer focus and it has the highest marketcap ever.

Herdr changed my tui game. Pi but for tmux.

Yea me too. I also wnet GUI for a little bit until I discovered herdr.

Can't you just run pi in normal tmux?

Also tmux has the worst tui I've ever used.

No but you can use bb or t3codes etc with it. Not perfect but allows orchestration between harnesses.

Yes, please allow me to have corps not only know everything about me but have a hyper accurate way to know what I’m thinking as well.

Or we could just type the whole sentence we need.

A friend made a keyboard for iOS because he didn't like any of the existing ones, and it doesn't send any data anywhere: https://typeneuma.com

Futo keyboard on Android has a transformer LM option that uses local ML for text prediction, I've found it to work very well. As far as I'm aware it's completely on device.

“Generated entirely on device” seems like an omission that doesn’t cover diagnostic, telemetry, “product improvement” logs.

It can be the same way as it works now, hash table with hits stats.