CP3405 Design Thinking 3 · TR2 2026 · Course Week 5

Sprint 5 — Integration & Intelligence Sprint

vW25 · Due Sun 21 June 2026 · 23:59 SGT Presentation · Mon 22 June 2026 Sprints 1–8 · TR2 2026
Sprint5 of 8
Release TagvW25
Tag DeadlineSun 21 June · 23:59 SGT
PresentationMon 22 June 2026
A1 DeadlineSun 28 June · 23:59 SGT 📓

What Sprint 5 Is About

Sprint 4 was your Software Development Sprint — you built the pipeline that collects data, runs it through LLMs, and seals a prediction. If you achieved that, well done. Sprint 5 raises the bar: this is the Integration and Intelligence Sprint, where your pipeline stops being a collection of scripts and starts behaving like a coherent, self-contained system.

The single question Sprint 5 asks is this: can your system run end-to-end with minimal human intervention? The Scrum team's job this week is to close the gaps — between agents, between scripts, between the human score and the final prediction file — so that by Sunday night, the only human decision is the Wild Card override.

Sprint 5 in One Sentence

Connect what you built in Sprint 4 into one coherent pipeline — from Friday market close to Sunday's sealed prediction — and make it visible in a dashboard or structured output your whole team can read.

Capability Tiers — Where Should Your Team Be?

Level 1

Manual

Data collected by hand. Predictions typed. No scripts. No automation.

Sprint 1–3 Baseline

Level 2

Structured

Scripts fetch data. Output saved to GitHub. Some automation in place but still manually triggered.

Sprint 4 → 5 Floor

Level 3

Automated Pipeline

GitHub Actions runs fetch automatically. LLM API calls scripted. Delta scoring runs after actuals are known. Prediction file structured and sealed by tag.

Sprint 5 Target

Level 4

Intelligent System

Full pipeline plus a live dashboard or structured HTML report your team can read. Prescriptive output, not just data. Human Score integrated into output automatically.

Sprint 5–6 Stretch

Sprint 5 Expectation

Teams that completed a working fetch_market_data.py and GitHub Actions trigger in Sprint 4 should be reaching Level 3 this sprint. Teams still at Level 1 need to close that gap this week — the window is narrowing as A1 approaches.

Sprint 5 Deliverables — What Goes into vW25

Your vW25 release tag must contain the following. Seal it by Sunday 21 June 23:59 SGT. Late tags cannot receive a delta score.

Deliverable Owner Level Description
Prediction file
prediction_W25.json
R7 R2 REQ SPX, NDX, IWM directional prediction + confidence + Human Score Wild Card. Structured JSON or Markdown. Sealed inside vW25 tag.
Agent outputs
Almanac · Macro · Technical
R3 R4 R5 REQ Each agent writes independently before LLM synthesis. All three must be present and committed before the tag is cut.
LLM synthesis
ChatGPT · Gemini · DeepSeek
R8 REQ Identical prompt sent to all three mandatory LLMs. Responses saved and committed. Comparison table showing where LLMs agree and diverge.
Automated fetch log R9 REQ L3 Evidence that data was fetched automatically (GitHub Actions run log, auto-committed data file with timestamp, or equivalent). Teams at Level 2 must show manual fetch with a commit timestamp before Friday midnight SGT.
Pipeline diagram or README update R1 R9 REQ A diagram (FigJam, draw.io, or Mermaid in README) showing end-to-end data flow from market close → prediction seal. R1 signs off the README update.
Delta score report — vW24 R10 R6 REQ R10 calculates direction accuracy and error % for last week's vW24 prediction vs Friday's actual close. Committed as delta_W24.md or equivalent. R6 provides actuals data.
Dashboard or structured report R9 R1 STRETCH L4 A simple HTML page, Streamlit app, or GitHub Pages view showing the current week's prediction, agent summaries, and LLM comparison in a readable format. Not required to be live — a screenshot in the repo counts.
Sprint Retrospective notes R2 SCRUM R2 Scrum Master documents what the team completed, what was blocked, and what changes in the next sprint. Goes into the Weekly Workbook and FigJam board.

Role Focus — Sprint 5 Priorities

R1 · Product Owner

System Coherence

Define what "done" means for the integrated pipeline. Sign off the README. Ensure all agent outputs connect to the prediction file.

R2 · Scrum Master

Integration Unblocking

Chase every role daily. Run a mid-week check by Wednesday. Retrospective is mandatory this sprint — A1 workbook needs it.

R3 · Almanac Agent

Seasonal Context

Historical patterns for the week of 16–20 June. Document sources. Output must be committed before R8 runs LLMs.

R4 · Macro Agent

Fed + Data Watch

Macro events for w/c 16 June: Fed speakers, CPI revisions, any major data releases. Structured Markdown output.

R5 · Technical Agent

Trend + Momentum

SPX/NDX/IWM technical read from Friday 13 June close. Key levels, trend direction, momentum signal.

R6 · Data / Actuals

vW24 Actuals

Pull confirmed Friday 13 June closes. Feed to R10 for delta scoring. Also prepare data structure for current week's collection.

R7 · Human Score

Wild Card Override

Review all three LLM outputs + three agent outputs. Apply override if consensus is wrong. Write Wild Card insight — this is the primary marks differentiator.

R8 · LLM Operator

Synthesis + Comparison

Send identical prompt to ChatGPT, Gemini, DeepSeek. Save all responses. Build comparison table. Claude optional if access available.

R9 · GitHub Lead

Pipeline Owner

Get GitHub Actions fetch running automatically if not yet done. Ensure all branches merged before Sunday. Cut the vW25 tag.

R10 · Calibration

Delta Score vW24

Calculate how accurate last week's prediction was. Direction correct? Error %? Commit delta_W24.md with retrospective notes.

Five Things That Will Sink Your Sprint

01

Waiting until Sunday to merge

If all branches land Sunday afternoon, there is no time to fix merge conflicts. R2 must chase commits by Thursday. R9 merges by Saturday.

02

Agent outputs after LLM synthesis

R8 cannot run LLM synthesis until R3, R4, and R5 have committed their outputs. The dependency is strict. R2 enforces the order.

03

No delta score for vW24

R10 must commit delta_W24.md this sprint. A prediction sealed last week with no retrospective score is an incomplete Scrum cycle.

04

Wild Card with no reasoning

R7's Human Score is the primary differentiator. "I agree with the LLMs" is not a Wild Card. Write a specific, sourced, non-obvious insight.

05

Missing the vW25 tag

A tag not cut by Sunday 23:59 SGT cannot receive a delta score next sprint. The tag is the immutable seal. There are no extensions.

Sprint 5 Timeline

Monday 15 June 2026

Sprint 5 Kickoff — after Sprint 4 presentations

Teams receive Sprint 5 brief. R1 and R2 plan the week's backlog.

Mon–Wed 15–17 June

Agent work week — R3, R4, R5 research and write

Almanac, Macro, and Technical agents commit initial outputs. R9 ensures pipeline is running.

Thursday 18 June

R8 LLM synthesis — runs all three mandatory LLMs

Identical prompt sent to ChatGPT, Gemini, DeepSeek. Comparison table committed. R10 begins delta_W24 calculation.

Friday 20 June — US market close

R6 collects actuals · Automated fetch runs

Friday close data for SPX, NDX, IWM collected. GitHub Actions pipeline should trigger automatically at 22:00 SGT.

Saturday 21 June

R7 Human Score + Wild Card — final review

R7 reads all outputs, applies override if warranted, writes Wild Card insight. R2 chases any missing role submissions.

Sunday 21 June 2026 · 23:59 SGT ⚑

vW25 Release Tag Deadline

R9 cuts the vW25 tag. All prediction files, agent outputs, LLM synthesis, and delta score must be committed before the tag is created. No extensions.

Monday 22 June 2026

Sprint 5 Presentations · Sprint 6 Kickoff

Every role (R1–R10) speaks. Demo of the integrated pipeline. R10 presents delta score for vW25 prediction.

📓

Assessment 1 — Weekly Workbook — Due Sunday 28 June 2026 · 23:59 SGT

Sprint 5 is the last sprint before the A1 deadline. Your Weekly Workbook covers Weeks 1–5 and must include: Knowledge Checks, Application Tasks, AI Conversation Reflections, and a Micro Reflection (~200 words) for every week. Sprint 5 activities — your pipeline diagram, delta score, agent outputs, and retrospective notes — all feed directly into the workbook. Do not leave the workbook until after Sprint 5 ends.