@vboykis Just wanted to add to this. LLMs these days are over optimizing things, because most if not all my technical conversations are like your call, you decide, that’s your decision.
Also cross referencing memories in chats is a terrible idea!
@vboykis The human brain cannot keep up with the AI token inference, and it’s getting faster. Sometimes I feel like humans are now becoming the bottleneck lol.
@naval By the time you’re at the last stage i.e. doing the actual work, the desire starts to fade, or perhaps a new desire has already taken its place.
@deedydas I think perplexity is living on those early contracts, and companies now waiting to get out ASAP
I don’t even know mercor’s business they like new scale ai collecting training data?
Truly a vibe economy!
@astnkennedy Yeah.
The thing is AI produces too much, too fast which I don’t think human brains can keep up, and if you try taking a look all the changes, then, you better off writing the code on your own. Perhaps the reason for these new vulnerabilities that we see almost every week now.
@karpathy@cosminnegruseri My outsourced thinking comes back at a blazing fast inference speed ranging from 10s to 100s of tokens/sec and that’s completely out of sync my brian’s ingestion speed (understanding).
First-ever RSE session at HPDC’26, see details below:
35th ACM International Symposium on High-Performance Parallel and Distributed Computing (HPDC) July 13–16, 2026, Case Western Reserve University, Cleveland, USA
HPDC 2026 invites submissions for RSE (Research Software Engineering) Lightning Talks as part of the conference program.
The RSE lightning talks provide a platform for sharing experiences, ideas, and insights related to research software in high-performance and distributed computing.
We encourage participants to present short talks on their work, tools, and experiences that contribute to the development and sustainability of research software.
For more information, submission guidelines, and deadlines, please visit: https://t.co/oOzHqN2UUc
#HPC #AI #ResearchSoftware #Memory #AISystems