Real-World Examples
See the Numbers
Concrete examples with actual token counts, costs, and checkpoint data. See exactly how much resumability saves.
95%
Avg Token Savings
98%
Avg Time Savings
$0.35/resume
Avg Cost Savings
2-8K tokens
Checkpoint Size
Scenario
A developer is reviewing a complex pull request that touches authentication, database queries, and API endpoints. The AI has analyzed the code, identified issues, and is mid-way through suggesting improvements.
Timeline
09:00 - Started PR review
09:15 - Analyzed auth changes (5 files)
09:30 - Found SQL injection risk
09:45 - Suggested fixes, awaiting response
14:00 - Developer returns from lunch
Hydration Approach
45K
tokens
~8 seconds
time
$0.135
cost
1.Load 47 conversation messages
2.Re-read all code snippets (12,000 tokens)
3.Reconstruct analysis context
4.Re-derive security findings
5.Rebuild suggestion state
6.Ready to continue
Resumability Approach
1.2K
tokens
~200ms
time
$0.0036
cost
1.Load checkpoint (1,200 tokens)
2.Restore cognitive state
3.Ready to continue
Savings Summary
44K
tokens saved
97.3%
reduction
$0.1314
saved per resume
Sample Checkpoint Data
json
1 { 2 "id": "cr-pr-review-001", 3 "version": "1.0", 4 "contentHash": "sha256:a1b2c3d4e5f6...", 5 "task": { 6 "id": "task-001", 7 "goal": "Review PR #847: Auth Refactor", 8 "status": "in-progress", 9 "successCriteria": [ 10 "No security vulnerabilities", 11 "Code follows style guide" 12 ] 13 }, 14 "decisions": [ 15 { 16 "id": "dec-001", 17 "description": "Recommend parameterized queries for User.findByEmail", 18 "status": "tentative", 19 "rationale": "String interpolation creates SQL injection risk", 20 "overridableBy": [ 21 "human" 22 ] 23 } 24 ], 25 "questionsForUser": [ 26 { 27 "id": "q-001", 28 "type": "question", 29 "question": "Should I also check for XSS vulnerabilities in the API responses?", 30 "priority": "medium", 31 "status": "open" 32 } 33 ], 34 "resumePoints": [ 35 { 36 "id": "rp-001", 37 "label": "Continue security review", 38 "when": "Developer returns from lunch", 39 "preconditions": [ 40 "SQL injection fix reviewed" 41 ], 42 "safeToAutoExecute": true, 43 "approvalStatus": "pending" 44 } 45 ], 46 "learnings": [ 47 { 48 "id": "learn-001", 49 "content": "This codebase uses raw SQL in several places", 50 "category": "pattern", 51 "confidence": 0.9 52 } 53 ], 54 "currentState": { 55 "phase": "execution", 56 "knownFacts": [ 57 "PR touches 5 files in /auth", 58 "User.findByEmail uses string interpolation" 59 ], 60 "uncertainties": [ 61 "Are there other SQL injection points?" 62 ] 63 } 64 }