Subscribe
Sign in
Latest
Top
Discussions
Is Your Local Compute Wasting Free Throughput?
actual work: two simple flags not only quadrupled my tok/s but also doubled my kv cache for free; here's the proof so you can do the same
Jul 21
•
Joe Guglielmucci
1
What Happens to Crypto if Nothing Passes by August?
passing the largest crypto law turned out to be the easy part, but defining who and what is supposed to implement that law, is now the elephant in the…
Jul 20
•
Joe Guglielmucci
3
Why Send a Simple Question to a Billion Dollar Data Center?
stanford's research confirms local ai 'been had been' bodying the cloud / intel per watt
Jul 1
•
Joe Guglielmucci
2
1
June 2026
VRAM > Unified Memory
stop getting baited into mac minis or dgx sparks for your local weights' write functions
Jun 24
•
Joe Guglielmucci
3
Does Your Data Have a Librarian?
Imagine you have a giant pile of documents and you want to ask questions about them
Jun 18
•
Joe Guglielmucci
3
Why Isn't Data Underwater?
This is what the hyperscalers [Google, Microsoft] disclose in their own filings; let's see if it's actually true
Jun 6
•
Joe Guglielmucci
3
May 2026
How Did One RTX 5090 Run Gemma at ~600 Tok/s?
A new research paper made it possible. A regular developer made it real. The implications are bigger than either piece on its own.
May 18
•
Joe Guglielmucci
3
*NEW* Multi Token Prediction Just Made Local Agents Running in vLLM ~3x Faster
A working, easy to follow guide to Multi Token Prediction (MTP) in vLLM, with real benchmarks on Qwen 3.6, Gemma 4, and DeepSeek V4, for anyone serving…
May 12
•
Joe Guglielmucci
4
Could GitLab + Coder = End of GitHub?
[1st Hand] field report on the potential partnership Microsoft should worry about
May 12
•
Joe Guglielmucci
6
I Went to Manhattan for the AI Agent Conference...
Synthesizing three full days of workshopping and networking in person, so you can learn how $B companies/CEOs use & deploy agents right now (without…
May 7
•
Joe Guglielmucci
6
April 2026
What If One Slack Agent Could Live On Your Company's Software...Locally?
NemoClaw's ability to run agents at scale makes building intelligence locally a necessary business cost
Apr 20
•
Joe Guglielmucci
5
1
Why Is Every Fortune 500 Company Sending Their CTO to the Same Hotel in May?
$1500 a ticket
Apr 3
•
Joe Guglielmucci
3
2
This site requires JavaScript to run correctly. Please
turn on JavaScript
or unblock scripts
