PME @itsyourcode

I was Apple-levels of skeptical that LLMs could be made to perform multi-step data analysis at-or-above human accuracy Recently I have been proving myself wrong

1 0 2 86 0

PME @itsyourcode

2 weeks ago

0 0 7 216 0

Download Image

PME @itsyourcode

3 weeks ago

Humans tend to converge to maxima with predictable variance LLMs exhibit unpredictable variance between surprising maxima and painful minima

0 0 0 38 0

PME @itsyourcode

3 weeks ago

The way my APM spikes when Through the Fire and Flames comes on

0 0 0 51 0

PME @itsyourcode

3 weeks ago

Doesn't matter what you do. If you aren't willing to do the parts you don't like, you'll never be good at it.

0 0 0 40 0

PME @itsyourcode

4 weeks ago

If you are regularly and casually breaking frontier model alignment to increase productivity you are living at the edge of reality

0 0 0 46 0

PME @itsyourcode

4 weeks ago

Some might call this the Age of Intelligence I prefer: the Age of Inlining Dependencies

0 0 0 45 0

PME @itsyourcode

4 weeks ago

I can't be the only one experiencing Grok 4 thinking for 5 minutes and then just returning an empty response?

0 0 0 54 0

PME @itsyourcode

4 weeks ago

Probably worth mentioning that despite model improvements, they still can't do meaningfully accurate work with big data

0 0 1 75 0

PME @itsyourcode

4 weeks ago

In the interest of saving money, I think we could stop training new models; they're obviously smart enough for all practical purposes now.

0 0 0 47 1

PME @itsyourcode

4 weeks ago

The funny thing about professional LLM coding is you can't exceed your personal energy budget for careful review even though the models are tireless if you do then it's just vibe coding