Sean Miller

I help companies scale profitability without scaling headcount ➡️ Building the Deloitte Digital for the non-Fortune 500 in the midwest

[GREG ISENBERG @gregisenberg]

Today is America's 250th birthday.

1 year ago, I became of citizen of the USA.

America is the one place where chasing your crazy idea makes people root for you. Whatever the idea is.

A startup nobody asked for. A pop-punk band at 30. A weird little product you can't stop thinking about.

You say it out loud here and people get excited and ask how they can help. That's the magic.

Other places I've lived, I've gotten torn town. Not here.

The USA celebrates you for going after the thing you actually want to build. 250 years in, the best thing America still gives you is permission.

Permission to try. Permission to reinvent yourself. Permission to believe you can do better than the people who came before you.

It isn't perfect. No place is. But I think it's worth celebrating.

Happy birthday to the best place on earth to build your dream.

I'm grateful to be part of it.

[Sean Miller @seanfrompulse]

@MichaelHyatt Claude Sonnet 4.5 all the way!

[Sean Miller @seanfrompulse]

@MichaelHyatt @claudeai The max context length has thrown me for a loop several times. Projects have been a great way to get around this for longer info dumps - have you explored these @MichaelHyatt?

[Sean Miller @seanfrompulse]

@noahkagan Meet has improved massively the past few years. Hands down the best product on the market and significantly cheaper. You’ll have no issues!

[david friedberg @friedberg]

free speech. but we’ll shoot you.
something is deeply broken in the west. if we don’t fix it, eventually everyone could be ushered at gunpoint to sign the proclamations of the Truth Ministers.

[Sean Miller @seanfrompulse]

@angelataptos @sama Need. more. energy.

[Sean Miller @seanfrompulse]

@aitomorroww @sama And yet, most of the world still hasn’t tried ChatGPT or heavily adopted AI into their workflow. I think there’s significant more compute needed before we scratch the surface of end-to-end agentic workflows at scale.

[Sean Miller @seanfrompulse]

@sama ~2 million GPUs signups + latent data center machines on https://t.co/3Xr58yASjP! H100s, A100s, and hundreds of thousands of RTX-series cards! Would love to see if we could help take your research and data processing pipelines off your main data center servers!

[Sean Miller @seanfrompulse]

@GregKamradt 92% accuracy for Assembly, 91.2% on Salad, but 50% cheaper on Salad ($0.12 vs. $0.06/audio hr). Accuracy seems to be a given for the top models.

[Sean Miller @seanfrompulse]

@ankurnagpal 👀

[Sean Miller @seanfrompulse]

@ElonMuskAOC What color are the messages?

[Daniel Sarfati @drsarfati]

“Why would someone use Salad?”

Salad is designed with scale in mind, if you only need a few mins of GPU Salad isn’t a fit. If you need to run 10+ GPUs 24/7 then give Salad a try. We take the work out of managing compute, just give us a container and we take care of the rest.

[Sean Miller @seanfrompulse]

@ashvinmelwani LP

[Sean Miller @seanfrompulse]

@thenicolaspizzo seo

[Sean Miller @seanfrompulse]

@EcomShri ceo

[Sean Miller @seanfrompulse]

@jspujji @ampush Bootstrap

[Salad Cloud @SaladTech]

How to deploy YOLOv8 on Salad's GPUs to detect objects in live streams: https://t.co/0MzTsGhDlG (Includes a step-by-step guide).

GPUs are a perfect fit for YOLOv8. Using GPU’s on Saladcloud, we can process each video frame in less then 10 milliseconds, which is 10 times faster then using a CPU.

On Salad, you can deploy YOLOv8 with GPUs starting from $0.032/hr. For comparison, the lowest GPU compute in Azure is $0.19/hr, 6X more expensive.

#computervision #yolov8 #objectdetection #ultralytics

[Salad Cloud @SaladTech]

NEW BENCHMARK: #stablediffusion v1.5 on consumer GPUs (up to 20,000 images per dollar)
Read here: https://t.co/MhdSzYBo2Q

Some Observations:
- You can get surprisingly good cost performance out of the 20-series and 30-series RTX GPUs
- Do not use the GTX series GPUs for production stable diffusion inference
- Newer GPUs with higher model numbers are faster in nearly all situations
- Batching saves time & money: 5-30% savings using batch size 4 vs 1
- Generation time scales close to linearly with number of pixels
- Stable fast gives significant cost savings while providing a fast and reliable image generation experience

#generativeai #gpu #benchmark #cloudcomputing

[Salad Cloud @SaladTech]

1000s of consumer GPUs lie unused most of the day.
100s of AI companies need GPUs to serve inference.

Salad connects the two: SaladCloud is a distributed cloud with 10k+ consumer GPUs at the lowest prices in the market.

Listen to our CEO, Bob Miles introduce SaladCloud and explain how AI/ML companies can finally stop overspending for GPUs.

Visit https://t.co/ytLABAIRbt for more information.

#GPU #CloudComputing #stablediffusion #GenerativeAI #MachineLearning #LLM #aitools