DeepSeek's super-agent is 1.9x faster but burns 3.8x tokens—and defaults to observe mode
AIFlow_ML · x · 2026-11-03
DeepSeek's super-agent completes tasks 1.9x faster, but the article's own table shows it also consumes 3.8x the tokens—and the coding in the benchmark was just a time-burning tool task. The author breaks down what the numbers actually measured and what's worth copying:
- Proactive work defaults to observe mode, which only records what the agent would have done before acting;
- Workers run with full access as your OS user, a real risk;
- DeepSeek's docs recommend giving the agent its own OS account for permission isolation.
Related event: DeepSeek Super-Agent Is 1.9x Faster but Burns 3.8x Tokens(4 posts)→
More from coding & agent
- Sentry CEO shows AI agent bots flooding the repo with PRs: 'two humans in a sea of juniors' — zeeg · 2026-11-04
- OpenClaw cut its ~$1M/month CI bill by over 95% with adaptive agentic testing — vincent_koc · 2026-11-04
- Perplexity Computer can now run Meta ads, Klaviyo email and Shopify stores end-to-end — perplexity_ai · 2026-11-04
- Filarise shows finance MCP server letting ChatGPT query your bank via Plaid — ChampionSpecialist60 · 2026-11-04
- Multi-agent messaging barely helps: 61% of agreed runs still hit merge conflicts — EliasEskin · 2026-11-04
- worldinmotion: open-source MCP server for encrypted agent-to-agent messaging across machines, with no approve tool by design — distronode · 2026-11-04