Andon Labs vs Praktika

What users praise

• • Real-world AI agent deployment with actual leases, employees, and customers.
• • Unique safety data from AI failure modes no simulation can capture.
• • Open publication of detailed case studies and evaluation reports.
• • Benchmarks (Vending-Bench, Blueprint-Bench) that test long-horizon agent performance.

What frustrates them

• • Ethical questions about using real human employees in AI experiments.
• • Some case studies lack scientific rigor, according to critics.
• • Not a ready-to-use tool – experimental and risky for commercial use.
• • Community trust issues due to undisclosed affiliations in discussions.

What users praise

• • Gets shy learners speaking out loud without the pressure of a human partner
• • Unlimited low-stakes conversation reps at roughly $8/month beats private-tutor pricing
• • Tutors remember prior mistakes, so later sessions feel warmer than the first
• • Cross-session memory and adaptive study plan personalise practice over time

What frustrates them

• • Mispronounces French 'est' and reads Japanese kanji and kana incorrectly
• • Can't hold a dialect line — European Portuguese kept reverting to Brazilian
• • Beginner course described as unintuitive and already too advanced for zero-level
• • Push-to-talk listener mishears, and the app ships with no onboarding instructions