Andrej Karpathy 'complains' of getting trolled on Twitter, his former boss Elon Musk responds; says: We need...
Andrej Karpathy is no stranger to Twitter , the social media platform now known as X. He's spent close to two decades there. But the AI researcher, who co-founded OpenAI and recently joined Anthropic , says the site has hit a new low. After a thread about Anthropic's Claude Tag feature spiralled into mockery and cheap shots, Karpathy aired his frustration with the platform's increasingly hostile culture—and pinned the blame squarely on X's recommendation algorithm. His former Tesla-era boss Elon Musk , who now owns X, didn't just hear him out. He agreed.
It started when Karpathy laid out the thinking behind Claude Tag , Anthropic's new workflow that lets Claude join a team inside Slack with access to chosen channels and tools. As someone working on the inside, he walked through what it took to build—the engineering needed to make it "just work" across tools, integrations, compute environments and memory. He called it the third major redesign of how people interact with large language models. The first paradigm was the LLM as a website you go to. The second was an app you download. This third one, he argued, is a self-contained, persistent entity with org-wide tools and context, working alongside teams of humans. Less like chatting with a bot, more like having a colleague who happens to write most of the code.
Andrej Karpathy says critics didn't read past the headlineThat framing didn't land with everyone. Plenty of users dismissed Claude Tag as little more than a chatbot wired into Slack, with some pointing to existing tools that already do something similar. Karpathy pushed back, saying many critics hadn't read past the headline before taking shots. "This is correct, I think a number of people on the tl didn't read past the title and made inferences and comparisons that are just wrong," he wrote. It wasn't a "feature" like some crappy Slack bot, he insisted, but an "org-level harness" whose difference would become clearer over time.
It started when Karpathy laid out the thinking behind Claude Tag , Anthropic's new workflow that lets Claude join a team inside Slack with access to chosen channels and tools. As someone working on the inside, he walked through what it took to build—the engineering needed to make it "just work" across tools, integrations, compute environments and memory. He called it the third major redesign of how people interact with large language models. The first paradigm was the LLM as a website you go to. The second was an app you download. This third one, he argued, is a self-contained, persistent entity with org-wide tools and context, working alongside teams of humans. Less like chatting with a bot, more like having a colleague who happens to write most of the code.
Andrej Karpathy says critics didn't read past the headlineThat framing didn't land with everyone. Plenty of users dismissed Claude Tag as little more than a chatbot wired into Slack, with some pointing to existing tools that already do something similar. Karpathy pushed back, saying many critics hadn't read past the headline before taking shots. "This is correct, I think a number of people on the tl didn't read past the title and made inferences and comparisons that are just wrong," he wrote. It wasn't a "feature" like some crappy Slack bot, he insisted, but an "org-level harness" whose difference would become clearer over time.
Next Story