Brian

ai, engineering & product @ maximalist & book lover.
sharing what I learn.

[Brian @brianrey_x]

@yuhasbeentaken Is still early and is hard to get all parts of the triangle. Speed, intelligence and cost.

There's always a trade-off

[Brian @brianrey_x]

@KushalVijay_ For how long?

[Brian @brianrey_x]

@ThisIsOthmane Why to make the chart reading the opposite way? Great outcome but is kind of confusing.

[Brian @brianrey_x]

@zack_overflow Actually is even free. Open code has great models for free.

[Brian @brianrey_x]

@rmdnmd @MakerThrive Ok, is a good way of stating it. But the problem that I see is that many times even a good prompt lead to not so good outcomes because many times the prompt vets buried in context so for long scope tasks it might not apply and you'd need a better model.

[Brian @brianrey_x]

@notnullptr The main limitation is retail hardware vs underlying LLMs technology. You can fit good models in memory but always the trade-off is how good they're at generalizing.

Llms have billions or trillions of parameters, it's a totally new amount of data.

Give it a few years...

[Brian @brianrey_x]

@WhiskyToad Keep trying and change your approach/ iterate.

Colonel Sanders started KFC in his 60s and Ray Kroc joined McDonald's at 52.

If you enjoy what you're doing, who cares how many times you fail? You've still got plenty of time/attempts.

[Brian @brianrey_x]

@itsvlady $200 Claude
$30 Grok for grokbot and some automations.
$50 Muse (I regret this one but I forgot to disable auto renewal)
$20 Chatgpt (not codex but chatbot)
$10 Open code Go (to test open source models)

I just like to try stuff but it's hard to manage all these subs XD

[Brian @brianrey_x]

@nicolas582d @ChatGPT Thanks! It's already live.

You can try it here in Chrome.

https://t.co/87R5DTkP2H

[Brian @brianrey_x]

@ChatGPT gives me great answers, but I forget half of what I read so I built a Chrome extension to highlight and annotate anything worth remembering inside the chat.

just submitted it to the chrome store and waiting...

would you use it?

[Brian @brianrey_x]

@AlbertoTICote No sé hasta dónde es sostenible.

En varios podcasts escuche que ahora la gente con ADHD que estaba acostumbrada a trabajar así tienen superpoderes. Pero no lo veo sostenible a largo plazo

[Brian @brianrey_x]

@Surendar__05 I love combination between hardware and software, battery autonomy and speed. I've used Linux over many years and I'm happy with Mac now. But some new Linux distros are getting really good.

[Brian @brianrey_x]

@vineerpasam Initial momentum, lack of friction and habits. Claude web chatbot came later than chatgpt and it produced the initial guau moment for many people.

[Brian @brianrey_x]

@ProductHunt My first browser extension to solve my pain and making it almost native.

https://t.co/87R5DTkP2H

[Brian @brianrey_x]

@MakerThrive @X do you get good results with grok for coding?

I've tried all versions since 4.5 and in long-context projects it struggles A LOT for me. at least in terms of code quality, taste, etc.

[Brian @brianrey_x]

@ChatGPT @AriesZhou @Joshua_WD it's alive!

[Brian @brianrey_x]

The worst part of using @ChatGPT is losing all the good stuff in old conversations so I built a simple Chrome extension to fix that.

You can now highlight and annotate useful answers without leaving the chat.

Just launched it on the Chrome Web Store!

https://t.co/87R5DTkP2H

[Brian @brianrey_x]

@themishra4402 distribution. it's really hard to change habits and once you get used to a platform and everyone is into it is hard to change.

i've tried to switch from whatsapp to telegram and signal many times and the problem again and again is...
network effects aren't strong enough.

[Brian @brianrey_x]

@ayebytes Siendo sincero, a dia de hoy claude tiene el mejor ROI en cuanto a calidad de código, uso y modelos. Especialmente con la nueva generación.

Yo tengo el mismo plan que vos y no encuentro mejor alternativa. ChatGPT es lo mejor que hay para el día a día y Claude para código.

[Brian @brianrey_x]

@AriesZhou @ChatGPT Yes, only on web version for now because it's easier to control/access data and yes it will persist in your browser storage even if you switch conversations.

Each highlight remember which conversation it belongs to!