AI factories generate intelligence in the form of tokens. The question is how many you can produce per watt.
NVIDIA DSX OS brings open, modular software to AI factories at scale—helping operators accelerate deployment, improve efficiency, and strengthen reliability across multi-tenant AI infrastructure.
Read the technical blog to learn more ➡️ https://t.co/ViUL9GhRAw
#NVIDIAGTC
CoPilot has a rubber duck mode: https://t.co/DeFGfkMmBN. Thanks @brunoborges for sharing this.
I do that all the time using agentctl https://t.co/ptVat8Yhx8.
"The @GitHubCopilot merge commit bloated the branch — it absorbed all of PRs #153, #154, and #155. The cleanest fix is to cherry-pick just the original feature commit" - @claudeai to the rescue
Claude is down, getting 500 errors across the board
API Error: 500 {"type":"error","error":{"type":"api_error","message":"Internal server
error"},"request_id":"req_011Ca22J6GNGmWMyY6KMwxBP"}
@Shpigford@claudeai yep, same error
API Error: 500 {"type":"error","error":{"type":"api_error","message":"Internal server
error"},"request_id":"req_011Ca22J6GNGmWMyY6KMwxBP"}
Its only mid day and already ran out of credits, glad I got the context persisted using SpecKit and CLAUDE.md, switching from @OpenAI Codex to @AnthropicAI Claude Code for rest of the day ....
AI coding works great… until you switch agents. That’s where context breaks, and everything gets reinterpreted.
I used SpecKit with a simple project-level governance & CLAUDE.md to enforce behavior.
Agents became interchangeable, hand-offs worked https://t.co/UaSJgrd2a7
"Our primary instrument may be dying, but our job is not dying. It is being elevated."
I've been living in Claude + Codex and a dumb editor like VS Code to review the code. That's been sufficient.