Hands-on with Manus, Devin, Operator, Claude Computer Use, Replit Agent, and how to evaluate real agent platforms.
The gap between custom agent frameworks and platforms — and why it matters.
How Manus plans, runs, and reports on multi-step tasks.
What Devin actually handles well on real repos, and where to babysit.
Browser-driving agents — capabilities, limits, and sensible first use cases.
Building computer-using agents on Anthropic's primitives.
Using Replit Agent for real projects, not just demos.
A practical eval rubric for picking an agent platform for your team.
Onboarding, handoffs, failure recovery, and the trust-building sequence.
