Another we usually learn with pair coding: problems that seem terrible up close are super silly from afar. If you can't find someone to describe it to, even if they know little or no code, imagining it works great.
This is hard to do on a deadline.
@viemccoy A considerable part of my efforts with software and stuff have to do with that fact, which I stumbled upon while practicing heavy meditation:
Reading the same book in phone and eink would produce two dramatically different results in my next sitting.
On second thought that probably has to do with why human consumption.
Why develop super expensive chemical defense system, hey plant, I take care of 200 of your cousins can you put together a few leaves.
I don't know how you guys are surprised by this.
Plants didn't painstakingly evolve a molecule for humans to aerosolize and inhale for the kicks, they developed a chemical weapon to fend off predators.
You are smoking pesticide.
This is not even a joke.
I have had incredibly high nicotine levels in my blood for the last 18 years and I NEVER get ticks.
I've been in the woods with non-nic-users who get a half-dozen ticks at a go, and I get zero because my blood is literally pesticide.
This is not even a joke.
I have had incredibly high nicotine levels in my blood for the last 18 years and I NEVER get ticks.
I've been in the woods with non-nic-users who get a half-dozen ticks at a go, and I get zero because my blood is literally pesticide.
@theo Going for a super long walk under the trees with the dogs, breathing fresh air, while discussing features with one model, while n other models implement is pretty good.
It also often results in good code. Turns out you need to think to code, and thinking requires O2. Who knew.
When I was a kid I had a book called "How to beat your Dad at chess", which basically taught common end game patterns which the reader was told to memorize. There was an interesting chess cognition argument to go along with it:
@analogalok I have almost that exact same rig I had no faith in Hermes but I'll have to try it I mean you can send prompts to gemma from your phone with tail scale anyway
@HlobodanShirosh It is very hard for a person like us, raised by a mysterious box unattended, ran by God knows who, to understand the value of a teaching that makes the most accomplished monks of the Himalayas weep.
@poopswag34 The same way you sit and stare at squiggly lines and get better at x subject, you can sit with one of many techniques like samadha, later you move that to the sensations involved in being with other people and body grows new brain cells to give you the hand that is needed
@guinnesschen Actually not a bad take. I do that quite a lot and is notoriously less bad than many alternatives.
Unlike humans machine only has this job, can hold a useful amount in their attention thingy.
With claude 20usd are about 15 questions.
Cursor has api gpt 5.4 nano on a second 20usd they throw for the kicks. So far ive HAMMERED that guy, and I still have like 400 M tokens to spend, and thats leaving composer 2.5 out, that I think runs kimi and that also is a workhorse. The web versiom is a bit less good but still lofe saver in a pinch