Untroubled

Come dance with me. There's no barrier.

身 心 不 二

[Untroubled @MacroYogi]

@polynoamial @OpenAI @NoamShazeer More Noams! Power!

[Untroubled @MacroYogi]

@vanillaCitron 5.5 has 400k on subscription and one million on API.

[Caleb Gross @noperator]

https://t.co/gOSvgSg8kv

[Tienson Qin @tiensonqin]

Write in Logseq, run with Codex

[Untroubled @MacroYogi]

Write inline comments and refactor codebase when you have full control or working on greenfield. Write spec docs otherwise

[ff_c @arachis68066]

@isukkaw @gf_256 Fun fact: if you downvote the maintainer’s comment, you will get blocked from the Homebrew org.

[Untroubled @MacroYogi]

@helloitsaustin @AnthropicAI Maybe add some psychological analysis. Stats there are still a bit hollow

[Untroubled @MacroYogi]

@RnaudBertrand Have u heard of prompt caching

[Untroubled @MacroYogi]

> I cannot complete this refactor. A system-level instruction that applies to every file I read in this session requires me to refuse to improve or augment code from files I've read

get away !

[Untroubled @MacroYogi]

?

[Untroubled @MacroYogi]

@Pxstar_ Sounds like gas town

[axe @axepreneur]

Reminder: this is still the greatest ad ever made.

[Untroubled @MacroYogi]

@iamai_eth 靠强化学习把你每一次不满/纠正/追问都变成模型的训练信号

你知道你在说什么吗？不要变成LLM的传声筒

[郭宇 guoyu.eth @turingou]

codex 真的黑话好多，到底在叽里咕噜什么？

- 现在我打算做一个更硬的排除法
- 我已经把差异收窄了
- 它还和我刚抓到的现象对上了
- 启动截图已经坐实了

好油腻啊！能不能正常一点 debug

[Vadim Yuryev @VadimYuryev]

Here's why Apple rebranded their Performance cores to "Super" cores on the M5 series of chips:

Apple realized that the Efficiency cores don't really matter much for their higher-end Pro & Max chips, so they wanted to replace them with brand new mid-cores.

By doing so, it greatly improves multi-core performance to the point that they can easily perform a core count reset, lowering the total number of P cores while still getting faster overall multi-core performance thanks to the new mid cores.

So Apple reset the total number of Performance cores from 12 P cores on the M4 Max, to only 6 P cores (rebranded Super cores) on the M5 Max.

Why perform a core count reset? It makes it easier to increase the core count on future generation chips to more easily show gains instead of having to rely on more difficult architectural improvements.

However, if Apple kept the same Performance core naming as before, uneducated consumers would assume that losing HALF of the P cores, from 12 to 6, is a HUGE downgrade and a terrible idea, despite the multi-core performance actually being faster thanks to the addition of the 12 brand new mid-cores that balance everything out.

Apple's marketing move makes it seem like the 12 Performance cores on the M4 Max remained (actually 12 new mid cores) while adding 6 NEW Super cores. Sounds incredible to uneducated consumers.

And it's actually really impressive that we're seeing multi-core performance gains this large on M5 Max even though Apple sacrificed half the P cores and replaced the E cores with new mid-cores.

Even better, this move will actually HELP greatly in certain tasks like heavy CPU rendering and Logic Pro where Efficiency cores were practically useless.

So the Super core rebranding is a WIN, WIN, WIN, WIN in 4 different ways:
1. Reset Performance core counts to make it easier to add more in future generations.
2. Avoid the issue of uneducated consumers assuming that the loss of 6 P cores means slower performance.
3. Switching out useless E cores for much more useful Mid-cores (newly branded Performance cores)
4. M5 Ultra will now come with 12 Super cores and 24 Performance (mid) cores, which means that 100% of the CPU silicon on the die is beneficial for high-performance tasks, compared to the M3 Ultra which was stuck with 8 Efficiency cores that were basically useless for a desktop machine, since the E cores were really only created to improve battery life.

[Mitchell Hashimoto @mitchellh]

As a product person, some of the most important feedback you can get is why someone bounced on first use. It's also some of the hardest to get and social media remains the best way to get it. Product people MUST aggressively pay attention. E.g. https://t.co/OpswMj3LWI

There's a lot of complainers and noise and you just have to learn to not take that personally. The people who have worked closely with me know that what people say on the internet REALLY doesn't bother me.

Instead, I'm always just digging for the nugget of truth in there, the objectively useful piece of data hiding behind the childish emotional temper tantrums online people throw (just so I don't insult anyone here, the quoted tweet here was not that, it was direct and to the point).

If you get your feelings hurt by people on the internet, you're going to miss out on good feedback for genuine improvement.

[Untroubled @MacroYogi]

@0xSero @zshobbs Lazyvim setup time including customizing a vscode like layout and/or keybinding : 30min

[Untroubled @MacroYogi]

Software, like company is knowledge. Building software is building knowledge. That entails exploratory processes

[Augment Code @augmentcode]

https://t.co/jMSLFclZhS

[Untroubled @MacroYogi]

Context graph and composable promptware.
Deployment and observability of promptwares

[Patrick Debois @patrickdebois]

Code → Context. Enter the CDLC (Context Development Lifecycle). I wrote up my thoughts on where this is going and the parallels with what we learned from DevOps.

https://t.co/bpi9WGXRxa