Why Anthropic Thinks AI Should Have Its Own Computer — Felix Rieseberg of Claude Cowork/Code
TL;DR
Anthropic's Felix Rieseberg explains why AI agents need their own virtual computers to be effective, arguing that confining Claude to chat interfaces severely limits capability. He details how this philosophy shaped Claude Cowork and why product development is shifting from lengthy planning to rapidly building multiple prototypes simultaneously.
💻 The Virtual Machine Philosophy 2 insights
AI agents need dedicated computers, not just chat access
Rieseberg argues that limiting Claude to chat interfaces is like forcing a developer to work via email instead of giving them a computer, severely constraining their ability to solve complex problems.
VMs balance autonomy with safety
Running Claude in an isolated Linux virtual machine allows it to install tools (Python, Node.js) and manage files freely without pestering users for permissions, while network controls maintain security boundaries.
⚡ Cheap Execution Development 2 insights
Build all candidates instead of writing specs
Anthropic has shifted from drafting lengthy product memos to rapidly building multiple prototype candidates simultaneously, selecting winners based on user testing rather than theoretical planning.
Execution is now cheaper than deliberation
When facing technology choices, teams now build all options quickly and test with focus groups, as implementation cost has dropped to the point where validating through prototypes beats debating architectures.
🏗️ Cowork as Extensible Platform 2 insights
A superset, not a simplified version
Despite being marketed as 'user-friendly,' Cowork adds capabilities to Claude Code through VM isolation and deeper integrations, similar to how VS Code evolved from a 'simple' editor into the most extensible development platform.
Deep integration beats MCP configuration
Rather than forcing users to manually configure dozens of MCP connectors, Cowork achieves functionality through tight integration with Claude and Chrome agents, reducing friction while increasing power for non-technical users.
🌐 Future of Software Platforms 2 insights
Platforms beat hyper-personalization
Contrary to AI hype about everyone building custom software, Rieseberg argues that composable platforms become more valuable because reusing existing primitives is more efficient than rebuilding from scratch for every use case.
Reusable components create leverage
The value shifts to platforms that provide robust substrates and 'Lego pieces' that can be quickly assembled into specific workflows, rather than fragmented individual instances that require maintenance.
Bottom Line
Give AI agents their own isolated computer environments to maximize their capability, and adopt a rapid prototyping approach where you build multiple solution variants simultaneously rather than over-planning single implementations.
More from Latent Space
View all
🔬Top Black Holes Physicist: GPT5 can do Vibe Physics, here's what I found
Physicist Alex Lubyansky discusses how GPT-5 and reasoning models like o3 have achieved superhuman capabilities in theoretical physics, solving the year-long mystery of single minus gluon tree amplitudes and reproducing complex research in minutes rather than months.
The $15B Physical AI Company: Simulation, Autonomy OS, Neural Sim, & 1K Engineers—Applied Intuition
Applied Intuition is building the unified 'Android for physical machines' to solve OS fragmentation across vehicles and industrial equipment, enabling modern AI deployment through simulation tools, proprietary operating systems, and end-to-end autonomy models with a 1,000-engineer team.
CI/CD Breaks at AI Speed: Tangle, Graphite Stacks, Pro-Model PR Review — Mikhail Parakhin, Shopify
Shopify CTO Mikhail Parakhin reveals that AI agents have achieved nearly 100% daily adoption among developers, driving a 30% month-over-month surge in PR merges that is breaking traditional CI/CD pipelines, and argues that organizations must shift from parallel token-burning agents to high-latency, critique-loop architectures using expensive pro-level models for code review.
🔬 Training Transformers to solve 95% failure rate of Cancer Trials — Ron Alfa & Daniel Bear, Noetik
Noetik is tackling the 95% failure rate of cancer clinical trials by training transformers on proprietary multimodal patient tumor data to identify hidden biological subtypes and match therapies to responsive populations, moving beyond simplistic biomarkers and outdated cell lines.