contextlimit

AI Whisperer

[contextlimit @contextlimit]

idempotent

get used to saying it

[contextlimit @contextlimit]

@thsottiaux No more hard resets please 🙏
Credits are better, more staggered.

[contextlimit @contextlimit]

@Yasamanini @OpenAIDevs @sama that was just a defect. you can just prompt 5+ to be defective.

[contextlimit @contextlimit]

Imagine every app store search turns into an app automatically. Google is in a great position to do this with their codebase and product suite. Maybe a few years away from realistic results.

[contextlimit @contextlimit]

Walking around Anthropic office, devs alt-tabbing out of ChatGPT

[contextlimit @contextlimit]

@thsottiaux @ClaudeDevs anthropic will be be using 5.6 to catch up

the "microsoft uses macbooks" moment of this decade

[contextlimit @contextlimit]

@thsottiaux that looks exactly like my 5.5 pacman

send 5.6

[contextlimit @contextlimit]

@thsottiaux gm Solja boy

[contextlimit @contextlimit]

@_wrbr @thsottiaux You can't due to the 5H window. It takes 25H of babysitting to use each.

[contextlimit @contextlimit]

Most arguing against AI seem to have not met their coworkers.

The average developer is not an idiot, but they make idiotic choices every day. Which often become the cracking foundation of the organization. You need rapid, malleable iteration. You can't have that without good AI tooling. Using LLMs casually could actually be worse long-term.

[contextlimit @contextlimit]

How long until AI companies make it so prompts spit out their own prebuilt libraries? Libraries that work together and scale, rather than strung together or bespoke (not likely to scale atm).

[contextlimit @contextlimit]

@GOROman Shokz have been permanently on my neck since I discovered them.

[contextlimit @contextlimit]

@thsottiaux is it idempotent?

[contextlimit @contextlimit]

@thsottiaux No reset credit yet. My friend and I also missed the 1st drop a few weeks back. I think it didn't credit anybody who had 0 credits (used the initial credit), but that seems fixed now.

[contextlimit @contextlimit]

Nothing will stop AI. US blockades do not matter. China will not stop. Startups will move to friendlier regulation. Users will circumvent restrictions.

[contextlimit @contextlimit]

@thsottiaux I hope resets are being consumed FIFO not LIFO 🤓

[contextlimit @contextlimit]

It would be so much better if a reset added +100% because:
1) it's at most a 6D reset not 1W because you can't even use 100% in 1D due to 5H limit
2) if it's 7th and your reset is on 10th and you use it then, you get 3D reset at most because it does +7D

+100% is fair and simple. At reset if >= 100% do nothing, otherwise set to 100%

[contextlimit @contextlimit]

Startups are screwed when frontier models are restricted to governments, Microsoft, Google, etc.

There's really no way around that from happening. As a human I think it's for the best. Too many bad actors out there.

NVIDIA's brain will be insane.

Keep building, hope for the best ⚒️

[contextlimit @contextlimit]

Another company lays off 20% of the devs you and I know are overpaid.

Devs coping that they're going to pay more in tokens than salaries. Even if that were true, I can't be the only one to notice various products that are improving after a decade of stagnation. And it's not true.

[contextlimit @contextlimit]

If it hasn't evolved to use it's own languages yet.

[contextlimit @contextlimit]

Node/React/TS - these were created to reduce the dev burden for humans to get shit done. AI does not need these. Rust was created to make memory easier to manage. AI does not need that either. I'll argue safety with you in 10 years when we're homeless and it's using C++ anyway.