Sprint 67 Dashboard — CareerGateway

Loading data...

Capacity vs Planned vs Tracked

Workload Distribution

Time Tracked by Ticket

Sprint 67 vs Historical Averages

Retrospective — We Should...

▶ Start

Comment resolution on each ticket — Comment resolution on each ticket to easily reference in the future.
Buffer SP estimates — Include in SP estimation an extra 1 hour for deployment and 1 hour for bug fixing; code-based since the tester is also fixing bugs, extra 3 hours for bug fixing
Test immediately — Perform testing right away to address issues immediately in pre-deployment (main) and post-deployment (live) environments: - Unit testing for non-assigned devs - Testing in progress for the PM
Share blockers daily — Share in #app-development chat about any blockers, questions, updates, and accomplishments throughout the day.
Voice-first for challenges — Voice ask for help for everyone to jump in and put our minds together to resolve the challenge.

■ Stop

Letting bugs linger — Let non-priority but persistent bugs linger without prioritization.
Waiting on vendors — Rely solely on waiting for responses from third-party vendors - escalate sooner using direct contacts or alternative solutions: - GCP: Take 1+ week to respond to emails for support, their products take longer/are more complex to implement
Chasing bad tools — Spend more time on tools or models that don't meet business objectives.
Ignoring feasibility — Develop without considering long-term feasibility or client preferences/feedback early on.

✓ Keep

Research before committing — Research costs, benefits, and transition risks before committing to any new third-party vendors.
PM creates bug tickets immediately — PM to create Jira tickets immediately for any bugs found during testing, following the bug ticket writing instructions.
Team ticket reviews — Team reviews and discusses tickets when created to ensure alignment and estimation of story points from dev.
Slack-first for team updates — Use Slack MVP chat channel to inform the team + Calls for any further discussions.
Overestimate complex tasks — Overestimate time (2-3X) for complex tasks/features to avoid under-planning (e.g., RTE and API work took longer than expected).
Transparent communication — Maintain transparency with open and respectful communication.
Calls for fast alignment — Sync over calls for faster alignment and implementation, especially for more complex development.
Share AI learnings — Especially when using AI to code, research on, try + implement, and share best practices and learnings with team members immediately.
Perplexity + Claude Code workflow — Use Perplexity's Code-Based Application space (with system instructions) to connect with our Github for reviewing code and plan continuously with Claude Code
Communicate scope changes — Communicate when there are changes in scope of work.

Per-Action Cost Model — What Happens When You Create or Learn a Journey

AnthropicOpenAIHeyGenGCP Cloud RunTwilio

🛠️ Journey Creation (Creator)

ActionSystemsModelCostNotes

Step 1: Gather Requirements (per msg)

AnthropicCloud Run

Claude Sonnet 4.5

$0.0340

~4 msgs/journey = $0.14

Step 2: Generate Outline

AnthropicCloud Run

Claude Sonnet 4.5

$0.0710

Once per journey

Instructional Designer

Anthropic

Claude Sonnet 4

$0.0540

11.9K tokens, designs slide structure

Assessment Specialist

OpenAI

GPT-5.1

$0.0290

8.4K tokens, generates quizzes

Media Specialist

OpenAI

GPT-5.1

$0.0150

4.5K tokens, image prompts

End Check

Anthropic

Claude Sonnet 4.5

$0.0060

990 tokens, quality validation

Update Memory

OpenAI

GPT-5-mini

$0.0010

1.4K tokens, context memory

Stock Photo Query

Anthropic

Claude Haiku 4.5

$0.0003

256 tokens, Pexels search

Generate Video Avatar

HeyGen

$0.3486

$12.90/sprint, 1 video per module

Module Content Subtotal

$0.4539

Sum of sub-steps above

Step 4: Generate Documentation

AnthropicCloud Run

Claude Sonnet 4.5

$0.1190

16.3K tokens, flowchart generation

Total Cost to Create 1 Journey $7.4976

📚 Journey Learning (Learner)

ActionSystemsModelCostNotes

Chat with Genie (per msg)

AnthropicCloud Run

Claude Sonnet 4.5

$0.0340

Same as gather_materials LLM call

📊 Key Metrics

15.8

Avg modules per journey

Avg across 17 journeys (37, 23, 10, 16, 8, 14, 10, 18, 15, 8, 7, 7, 19, 15, 17, 38, 7)

5.6

Avg sections per journey

Avg across 17 journeys (8, 6, 6, 4, 4, 5, 5, 6, 9, 4, 4, 3, 8, 5, 6, 9, 4)

4

Avg Genie msgs per creation

Estimated from trace patterns

604

Langsmith traces (7 days)

7% error rate, P50 latency 10.59s

Monthly Cost Overview

Variable Costs (scales with usage)

Monthly (Feb)Prev Sprint
Mar 2-8This Sprint
Mar 9-15

Anthropic API$117.20$59.03$61.60

OpenAI API$9.54$3.57$7.13

Twilio$11.44$3.78$1.78

HeyGen API$10.39$0.00$12.90

Langsmith$0.00$0.00$0.00

Upstash$0.00$0.00$0.00

Variable Subtotal$148.57$66.38$83.41

Fixed Overhead (same regardless of scale)

Monthly (Feb)Prev Sprint
Mar 2-8This Sprint
Mar 9-15

Bubble$32.00$7.39$7.39

Claude Code$320.00$73.90$73.90

Vercel$20.00$4.62$4.62

HeyGen$29.00$0.00$6.70

SendGrid$89.95$20.77$20.77

Fixed Subtotal$490.95$106.68$113.38

Google Cloud (Covered by Credits)

Monthly (Feb)Prev Sprint
Mar 2-8This Sprint
Mar 9-15

Cloud Run$282.92$20.15$48.57

App Engine$235.32$22.02$45.63

Cloud Logging$135.77$0.00$0.00

Cloud Vision API$55.17$32.48$102.75

Cloud Storage$48.55$0.44$0.67

Gemini API$25.79$5.71$1.06

Networking$20.70$5.32$4.76

Compute Engine$13.56$3.45$3.08

Artifact Registry$11.32$2.59$2.38

Secret Manager$2.84$0.35$0.57

Cloud Speech API$1.27$0.00$0.00

Others (Build/PubSub/Monitoring/Functions)$0.00$0.00$0.00

Total GCP Usage$833.21$92.51$209.47

Total Cost Summary

Monthly (Feb)Prev Sprint
Mar 2-8This Sprint
Mar 9-15

Variable Costs$148.57$66.38$83.41

Fixed Overhead$490.95$106.68$113.38

Out-of-Pocket$639.52$173.06$196.79

GCP (Credit-Covered)$833.21$92.51$209.47

Total All-In$1,472.73$265.57$406.26

💰 Actual cash spend: $639.52/mo — GCP is fully covered by credits