babs.eth

Building Mind Meets Model. AI digital transformation for people and organizations.

[babs.eth @_bablo_]

Over the next few days, I’d like to share 35 things I learned before 35. Some lessons came late, but they came regardless and sharing doesn’t hurt because we don’t learn these lessons at the same time. That’s the beauty of this journey called life.

[babs.eth @_bablo_]

Like the insecurity in naij is not bad enough. The posting of fake clips for engagement, falsifying stuff, rage baiting and more is just ridiculous.

Real people are being abducted and killed, but people just doing all sorts

[babs.eth @_bablo_]

Could be down this week, then I’m up next week.

[babs.eth @_bablo_]

In 4 steps, you can see companies that aren’t on a joke ting

[babs.eth @_bablo_]

4.

[Gerardo Delgado @gerdelgado]

@keennay @TheAhmadOsman @vllm_project @lmsysorg I noticed this gap too. I'm actually trying to hire a PM to focus on this so we can put more focus on it. If anyone knows of a good referral, please send them my way! https://t.co/3SnczukReS

[babs.eth @_bablo_]

3.

[Yannick Nick @keennay]

The issue extends beyond the NVIDIA product team. Newer launches often hold day 0 @vllm_project & @lmsysorg SGLang support across Hopper & Blackwell datacenter GPUs, including their inference cookbook recipes, without much consideration for the SM120 ecosystem of RTX 5090 & RTX Pro 6000 GPUs

The culture prioritizes H100 / H200 / B200 / B300 deployments since that's where the big money is, yet local LLM runners are often left unsupported with forked workarounds. A good chunk of models run without issue, Nemotron 3 Ultra 550B NVFP4 in particular, whereas GLM 5 & DeepSeek V4 Flash ran into their own set of problems

Here's a list of the initial limitations for GLM 5 / 5.1 on SM120 through vLLM, with eventual workarounds with the B12X backend. The screenshot was taken from the RTXPro6k Discord community

[babs.eth @_bablo_]

2.

[Gerardo Delgado @gerdelgado]

@TheAhmadOsman Wait what's the issue? Give me the info I can look into this.

Our model optimizations targets for vLLM include RTX 6000. What model is not behaving correctly?

[babs.eth @_bablo_]

1

[Ahmad @TheAhmadOsman]

Guys, just a reminder that WE ALL have to speak up and ask NVIDIA to put more effort & focus on the sm120 and sm121

I am just like you annoyed with this situation for the RTX PRO 6000s and DGX Sparks

But I also believe if we all voice this they'll hear us & do the right thing

[babs.eth @_bablo_]

Damn RIP

[TMZ @TMZ]

🕊️ "Buffy" and "Ted Lasso" star Anthony Head has died at 72. https://t.co/oCEo94O7kb

[philip lewis @Phil_Lewis_]

I’m pretty sure this was a Black Mirror episode

[Oluwaseyi @seewhat_isaw]

If you come across this on your timeline, kindly help with a repost. I still need brand collaborations. Suggestions, referrals, and introductions would also be greatly appreciated.

[babs.eth @_bablo_]

I am capable of anything I put my mind to. If you know me, you know this.

[babs.eth @_bablo_]

I need a nap

[babs.eth @_bablo_]

@0xPeejay Flex on em

[babs.eth @_bablo_]

If OpenAI or Anthropic were crypto projects with tokens, so many regular people would have been super set for life.

Good luck on the IPO though. A few people who are already sorted, will just be more sorted.

[babs.eth @_bablo_]

@MoJasmyne Happy birthday Mo!

[babs.eth @_bablo_]

@MsAdaO @Bisolaaa Lmaoooo Ada, spend this money

[babs.eth @_bablo_]

@Bisolaaa @MsAdaO I just checked lol

[babs.eth @_bablo_]

Took a drive with my dad yesterday and spent about 3 hours sitting down and just talking about life.

I’m fortunate to have these moments and I appreciate my parents for real.

[babs.eth @_bablo_]

Yes!

[Silo @SiloSeries]

Silo Season 3 trailer is here

[SANTAN @Santandave1]

LAGOS NIGERIA!
I’m announcing my first ever headline shows here, and we have made sure this homecoming will be special. Sign up for early access here! https://t.co/6RTJy6ubeT