Andrew Thompson

Deep Learning researcher. Deep Learning / Model Architecture & Training

HF:

[Andrew Thompson @andrewthomxp8u]

@amazingdeals27 @iem_k_run Used it to speed up dev, catch bugs, and sanity-check the math. Tried the top models from Claude and OpenAI, but only Gemini actually spotted the errors and helped optimize the architecture.

[Andrew Thompson @andrewthomxp8u]

@zefa_start @StatsWire Xhigh haiku vs HIGH gemini

[Andrew Thompson @andrewthomxp8u]

@amazingdeals27 @iem_k_run I'm using two Gemini AI Ultra accounts with a 20x limit.

[Andrew Thompson @andrewthomxp8u]

@Leapofcode @sashikantsingh_ Luna killed haiku*

[Andrew Thompson @andrewthomxp8u]

@archiegarg__ @atomic_chat_hq for 12x price?

[Andrew Thompson @andrewthomxp8u]

@StatsWire In arena

[Andrew Thompson @andrewthomxp8u]

@StatsWire I mean opus 4.6 was higher fable 5.1 in coding.

[Andrew Thompson @andrewthomxp8u]

@bidyut_io @Techorbitt Waiting for facts.

[Andrew Thompson @andrewthomxp8u]

@Umesh__digital @vibeonX69 Agree

[Andrew Thompson @andrewthomxp8u]

@Gaus450 @Taniyatweets_ Claude is not "BEST"

[Andrew Thompson @andrewthomxp8u]

@AdinaAragon @shariar_design Agree

[Andrew Thompson @andrewthomxp8u]

@iamsaniydv @grok Chrome.

[Andrew Thompson @andrewthomxp8u]

@Saas_addy True. I use Arch btw

[Andrew Thompson @andrewthomxp8u]

@lrislzc You’ve got AI. You’ve got hands. You’ve got the internet. Yet you’re still just a fat piece of shit sitting there making posts.

[Andrew Thompson @andrewthomxp8u]

@QuinnyPig @DynamicWebPaige 09.10

[Andrew Thompson @andrewthomxp8u]

@ship_together Wrong. Kimi is distilled shit. A second-rate Claude. Same goes for GLM.

[Andrew Thompson @andrewthomxp8u]

@SidharthVijay_ You added some moronic prompt at the top and think it's funny. Haha, I'm literally rolling on the floor.

[Andrew Thompson @andrewthomxp8u]

@eapyk @mayuri_3015 I hate Nvidia. You lose

[Andrew Thompson @andrewthomxp8u]

@ParthiSub @mayuri_3015 I hate Nvidia. You lose