Co-Founder & CEO @NexlayerAI | Building the agentic cloud for autonomous software.
Atlanta, GA
Joined July 2008
- Tweets3.6K
- Following5K
- Followers2.2K
- Likes5.5K
Sal Stabler retweeted
Roadster nitter.gq/i/broadcasts/1yKAPwNgg…
Bring one real workload.
We prove the fastest setup that holds it. Your model. Your data. Dedicated US. Nobody else on those machines.
You only commit after it holds.
Then the speed is contracted. The bill is one number. No token meter. We run the stack.
nexlayer.com/dedicated
5
60
The model can change. The engine can change. The compute can change. The performance objective doesn’t.
Lock the SLO. Not vLLM, not the GPU, not the vendor.
Nexlayer Inference — dedicated capacity continuously optimized around the workload.
nexlayer.com/resources/blog/…
3
141
TOKENMAXXING IS DEAD.
Maybe metering intelligence by the token should die with it.
liquidbrain.ai
41
5 trillion parameters is insane.
What’s even more insane is trying to run it with today’s inference architecture.
We built something different.
#5trillion #oss #llm @ByteDanceOSS
LiquidBrain.ai
17
AI agents are moving from answering prompts to working continuously—for hours, days, weeks, eventually months.
When that happens, does paying for every token still make sense?
We think the token meter eventually disappears.
If not tokens, what becomes the unit of value?
0%Reserved AI capacity
0%Flat-rate compute
0%Token pricing survives
0%Autonomous compute market
0 votes • Final results 1
9
Sal Stabler retweeted
shipping: liquidbrain.ai
Unlimited tokens.
Unlimited context.
One fixed monthly bill.
live in 14 days → liquidbrain.ai
#liquidbrain #ai #infrastructure
Per-token pricing is going to look insane in a couple months.
The whole industry is optimizing the wrong variable. Everyone’s racing on model quality while the actual ceiling is that you can’t afford to run the best open source models at the scale real work needs.
Unlimited tokens, unlimited context, one flat bill will do more for what gets built than the next model release. I’ll say it louder for the people betting their roadmap on the meter staying: it’s not.
Tell me I’m wrong.
#unlimitedintelligence #infinitecontext #unlimitedtokens #unlimitedcontext #llm #opensourcellm
1
37
Most open‑source apps break in production. We graded 250 of them. The results are brutal.
relaxed-weasel-america250.cl…
#America250Apps, #OSSLeaderboard
#ProductionReady, #InfraScore #GitHubProjects
34
Most people think deploying complex open source software still requires a DevOps team.
For America’s 250th, we tested that assumption with 250 real apps. relaxed-weasel-america250.cl…
Here’s what happened:
If agents can now handle the infrastructure complexity of real open source projects at this scale, what else are we still doing manually?
The future is agent-native.
1
10
This is our small tribute to 250 years of American ambition.
🇺🇸 250 years.
🚀 250 apps.
☁️ One AI-native cloud.
The Nexlayer 250 is live.
Happy birthday, America.
Keep building. 🇺🇸
#250apps, #America250, #OpenSource, #AIagents #AgenticCloud #agenticinfrastructure #AICloud
14