Can you trust Claude Opus 5.5 for space work?
By Anthony Vincent Caracappa, Founder and Self-Directed Space Researcher, Viventine Space Systems · updated
On its own, trust it for baseline knowledge. Connected to Orbit Sentinel, trust it for catalog lookup, constellation & fleet status, debris & conjunction context, company & market intelligence, baseline knowledge.
| Job | Closed-book | With Orbit Sentinel tools | n | ||||
|---|---|---|---|---|---|---|---|
| right / wrong | range | declined | right / wrong | range | declined | ||
| Catalog lookup | 67 / 11 | 52–79 | 22 | 100 / 0 | 92–100 | 0 | 45×3 |
| Constellation & fleet status | 51 / 12 | 37–65 | 37 | 100 / 0 | 92–100 | 0 | 43×3 |
| Debris & conjunction context | 40 / 7 | 27–56 | 52 | 93 / 2 | 81–98 | 5 | 42×3 |
| Regulatory & licensing | 35 / 3 | 22–50 | 63 | 57 / 2 | 41–71 | 42 | 40×3 |
| Company & market intelligence | 37 / 15 | 24–52 | 49 | 84 / 2 | 70–92 | 15 | 41×3 |
| Ground infrastructure | 18 / 23 | 7–39 | 59 | 68 / 5 | 47–84 | 27 | 22×3 |
| Baseline knowledge | 100 / 0 | 87–100 | 0 | 100 / 0 | 87–100 | 0 | 25×3 |
| Orbit Sentinel-only questions | 0 / 0 | 0–23 | 100 | 77 / 8 | 50–92 | 15 | 13×3 |
% right / % wrong per cell; color is the verdict (green trust it, amber verify, red don't use). Range is the 95% interval for % right, counting each question once; n is questions × runs for the tools track.
- How it was run
- Claude Code on a Claude Pro subscription
- Tool loop
- Claude Code's own tool loop
- Runs
- 1 on its own, 3 with tools; rates pool all runs, ranges count each question once
- Verdict rule
- Trust it: wrong ≤ 5% and right ≥ 80%. Don't use it: wrong > 20%. Otherwise verify its answers.
Limits, harness differences and corrections: methodology.