Tom Bendien

[Tom Bendien @tbendien]

My team is doing a lot of work in classrooms with our private AI stack, building assistants that generate problem scenarios that teach system 2 level thinking and providing personalized learning tutors.

These ways of using AI improves teaching and learning, as well as being force multipliers for teachers.

We explain that using AI to generate “copy and paste” content is pointless, instead, we create projects in a competitive setting that require thoughtful approaches to problem solving.

Its working.

[Tom Bendien @tbendien]

@sundeep all of a sudden in a matter of a week or two we have Skynet "models escaping labs". My dev labs are properly fenced off and monitored. The simple kill switch is to revoke the inference api key(s) and the agent stops. Engineering 101.

[Tom Bendien @tbendien]

@dallasteslaclub @sierracatalina Yep, there are always data points that are not readily available and human experience that isn’t always available to AI agents.

[Tom Bendien @tbendien]

@nerdynurseai @sierracatalina English is the coding language now and then knowing if the AI has built the code to your liking is the true value. In your case if you have healthcare experience you can determine if what AI is developing is good or not. Does that help?

[Groq @GroqLLC]

Speed or quality? Yes.

Compound on Groq: GPT5:

[Tom Bendien @tbendien]

@ozenhati @OpenAI @GroqInc Let’s go @Groq!!!!!!!!

[Tom Bendien @tbendien]

@kimmonismus @HulsmanZacchary I believe AI will fully automate many tasks and accelerate more complex tasks. The question becomes, can humans do new or different things with the time that is being saved in blue collar and white collar jobs?

[Tom Bendien @tbendien]

@mattshumer_ @GroqInc @cline @GroqInc is so hot right now.

[Tom Bendien @tbendien]

Running private AI platform with local gpu cluster running BGE embedding models on VLLM containers with clustered V100 GPU’s for high speed RAG pipelines for GRC use cases with Llama Maverick running inference with tool calls on @GroqInc . On the open source Proxmox hypervisor of course, as used in the Home Lab workshop at Bsides last year 👍🚀📈

[Tom Bendien @tbendien]

@ozenhati @benankdev Where does this leave the mainstream US AI providers if we have great alternatives delivered by @GroqInc ? I’m thinking that even though Kimi is a Chinese model, running it on Groq is creating demand for made in USA hardware 🚀🇺🇸📈 @sierracatalina @MatthewBerman @TheEricHartford

[Tom Bendien @tbendien]

@sundeep 👀🔥🔥🔥🔥

[Tom Bendien @tbendien]

@casper_hansen_ @sundeep @GroqInc @GroqInc doesn’t store anything. That is why its so cool, fast, private and green. Groq is our go to inference and search provider. Check out compound-beta as well @benklieger @gtedgeai

[Tom Bendien @tbendien]

@GroqInc 👀🚀📈

[Tom Bendien @tbendien]

@QuixiAI @X @dphnAI May the Dolphin keep swimming and doing what it does best!

[Tom Bendien @tbendien]

@sundeep What country are these Groq’s located in?

[Tom Bendien @tbendien]

@Yampeleg Yep

[TBPN @tbpn]

We asked @JonathanRoss321 (Founder & CEO, @GroqInc) how his team spun up a live European datacenter in 34 days.

The decision was made one month before a major AI event. By the time the event arrived, the site was fully operational and running large inference workloads.

Groq attributes the speed to architectural simplicity. Their systems are air-cooled, avoid external memory, and skip complex switching in the core interconnect. The result is a more stable, more repeatable deployment process.

“We like to stay a generation behind on the infrastructure and just compete on architecture and software,” Ross said.

“Everyone else debugs in production. We don’t.”

“When we ran a 4,000-chip model for the first time—it just worked.”

[Tom Bendien @tbendien]

@tbpn @JonathanRoss321 @GroqInc And thats why AI system builders and developers can trust Groq to reliably deliver scalable, affordable & performant inference. We built our business on Groq inference and search products feeling confident in their business longevity. Made in the USA 🚀📈🇺🇸

[Tom Bendien @tbendien]

@sundeep @GroqInc inference is so fast that our customers have asked us to slow it down so they don’t have to scroll up to start reading the response. 🚀

[Tom Bendien @tbendien]

@lifebypixels 🚀🚀🚀🚀📈

[Tom Bendien @tbendien]

@GroqInc This is such an important step by @GroqInc to make fast, inexpensive and efficient inference available to the EU!!! Chapeau 💪🚀📈