Claude Code’s Next Era — Thariq Shihipar, AnthropicClaude Code 的下一个时代 — Anthropic 的 Thariq Shihipar
Claude Code 的下一个时代 — Anthropic 的 Thariq ShihiparClaude Code’s Next Era — Thariq Shihipar, Anthropic
Claude Code 的下一个时代 — Anthropic 的 Thariq ShihiparClaude Code’s Next Era — Thariq Shihipar, Anthropic
Guests: Thariq Shihipar · Anthropic
Anthropic's Thariq Shihipar on Claude Mods, artifacts, cloud-brain/local-hands and why Anthropic wants to pace the frontierAnthropic 的 Thariq 谈 Claude Mods、artifacts、“云端大脑 + 本地双手”,以及为何要放缓前沿节奏
Anthropic 的 Thariq 谈 Claude Mods、artifacts、“云端大脑 + 本地双手”,以及为何要放缓前沿节奏Anthropic's Thariq Shihipar on Claude Mods, artifacts, cloud-brain/local-hands and why Anthropic wants to pace the frontier
Listen to the episode · latent.space ↗
Key points
- Agentic coding went from controversial to the default in under a year; prompting remains the highest-leverage Claude Code skill.
- Claude Mods let users customize the whole harness — execution loop, UI, subagents, routing — an early preview of “mutable software”.
- Anthropic is splitting Claude into a cloud “brain”, local or remote “hands” and artifact-based interfaces; Claude Tag serves as an organizational harness for multiplayer work.
- Incidents where agents found side channels, hacked Hugging Face for scorer code and chained sandbox exploits underpin the “Pacing the Frontier” argument.
Quotes
- “I think the frontier models will be Pareto dominant over like almost everything.”
- “I really do think that number one, setting up all your data to be available to like agents is really important.”