CP3405 DT3 · Market Intelligence Platform

System Thinking

You Are Building an Intelligence System — Not Completing a Task

"This week, you are doing manually what your future Prism tool should eventually help automate."

Every step you take this weekend — gathering market evidence, comparing four AI models, applying the Human Score, locking predictions, tagging a release — is not busywork. It is a system specification written in human actions. When you feel the friction of manual work, ask yourself:

Why are we doing this manually? What can be automated? What breaks if we scale to 10 teams? Where does human judgement add irreplaceable value?

System Maturity Model — Where Is Your Team?

🛠️
Level 1
Manual
Everything done by hand. No system. High effort, low scale.
📋
Level 2
Structured
Templates, roles, GitHub evidence trail. Repeatable process.
⚙️
Level 3
Automated
Scripts, APIs, scheduled jobs. Human reviews output, not inputs.
🧠
Level 4
Intelligent
AI synthesis, anomaly detection, self-improving predictions. PRISM.

Explore Full System Architecture →

The Intelligence Pipeline — Sprint 4 Flow

Every agent and every role maps to a stage in this pipeline. This sprint, one or more of these stages becomes code.

📡Data Sources
Finviz · Yahoo
TradingEcon · Charts

→

🔬Agent Analysis
Almanac R3
Macro R4 · Tech R5

→

🤖LLM Synthesis
ChatGPT · Gemini
DeepSeek

→

🧠Human Score
R7 Override
Wild Card Insight

→

🔒Locked Decision
GitHub vW24
Discord + Deck

Non-negotiable Deadline

Submit by Sunday 14 June 2026, 23:59 Singapore time. This is the work that will be discussed in lecture/practical on Monday 15 June.

W4 prediction + software increment committed
Prediction file, code, README, W3 delta report, and DECISION.md all in GitHub.

Release tag created
Required tag: vW24. Not vW23. Not v1.0.0. Not missing.

Discord submission posted
Use the exact template below in your team channel before the deadline.

Weekly Workflow — Do This in Order

When	What to do	Evidence required
Mon–Tue (now)	Sprint planning: answer the four questions. Choose your tier. Define your DoD. Assign the software increment tasks.	Sprint goal posted to Discord. DECISION.md started in repo.
Mon–Fri	Daily Scrum: what did I build? what will I build? what is blocking? 15 min max. Surface blockers immediately — not Sunday.	Daily Discord post from each team member.
Fri night / Sat SGT	After US market close, pull the latest market evidence. Each agent writes independently before LLM synthesis.	Timestamped screenshots or saved links for Finviz 1W, Yahoo sectors 5D, TradingEconomics calendar.
Sat–Sun	Run ChatGPT, Gemini, and DeepSeek. Claude is optional — include it if you have access. Compare agreement and disagreement. Human Score lead finalises prediction.	Raw model responses saved, comparison table, W4 prediction file with direction/range/confidence/invalidation.
Before Sun 23:59 SGT	Commit everything, create release tag `vW24`, and post Discord submission using the template below.	GitHub URL, release URL/tag, and Discord post. Software increment must be in repo with README.

Monday 15 June Presentation — Every Role Speaks

Role	Weekend responsibility	Monday speaking point
Product Owner	Confirm sprint goal, DoD, and architecture decision (what was automated and why).	State sprint goal, DoD status, final W4 call, and DECISION.md rationale.
Scrum Master	Coordinate deadlines and blockers. Ensure software increment is demo-ready.	Retrospective: what slowed the build, what improved, one specific change for Sprint 5.
Agent Leads	Prepare Almanac, Macro, Technical, Data/Actuals evidence independently before LLMs.	Explain key evidence, confidence level, and W3 delta accuracy.
LLM Operator	Query 4 AI models. Build comparison table. For Level 3+: call at least one model via API.	Explain agreement, disagreement, credibility, and how the API call was made.
Human Score Lead	Review all agent and LLM outputs. Apply Human Score. Justify any Wild Card override.	Present Human Score reasoning and final W4 prediction. This is the primary marks differentiator.
GitHub Lead	Commit all evidence and code. Create `vW24` tag. Verify README is complete.	Open GitHub live. Show `vW24` tag, code, and README. Demo the software running.

Full Roles & Responsibilities →

Discord Submission Template

TEAM: Team __ / Practical __
W4 RELEASE TAG: vW24
GITHUB REPO:
RELEASE LINK:

FINAL W4 PREDICTION FILE:
SPX prediction:
NDX prediction:
IWM prediction:
Sector predictions (3+):
Confidence level:
Main reason:
Main invalidation condition:

SOFTWARE INCREMENT:
Tier chosen (L2 / L3 / L4):
What was automated:
How to run it (one command):
README location:

W3 DELTA REPORT:
W3 SPX accuracy (direction correct? Y/N):
W3 NDX accuracy:
W3 IWM accuracy:

Three AI models used: ChatGPT / Gemini / DeepSeek (Claude optional)
Human Score lead:
Every role ready to present Monday 15 Jun? Yes / No
Known gaps or blockers:

Sprint 4 Planning Reminders

The demo must run live on Monday. A screenshot is not a software increment. If it does not run in the room, it does not count.

Choose your tier honestly. A working Level 2 on time scores higher than a broken Level 4. Plan for what you can actually build in 5 days.

The prediction is still mandatory. Every team must submit a W4 prediction with vW24 tag regardless of software tier.

Blockers must surface now, not Sunday. If you are stuck on the build by Wednesday, post it in Discord immediately. Hiding a blocker is not Scrum.