MICROSOFT GLOBAL HACKATHON 2026
Every Windows PC has Start. I chose that familiar entry point to explore a bigger role: not just opening apps, but helping people move their work forward.
One week. An existing product. A better way to work.
Morning handoff
Draft ready · 1 decision needs you
FROM THE OVERNIGHT REVIEW
Keep option B. Check larger text.
Teams US review + Figma Aster feedback
For Maya · before your 11:30 critique
Review update
Why this?
SPRINT
1 week / hackathon exploration
MY CONTRIBUTION
Framework, learning session & Start prototype
STATUS
Concept / not a shipped Windows feature
THE HACKATHON BRIEF
The brief was to improve a work process on an existing product. I had one week to make the idea tangible—not to redesign all of Windows.
Start was the opportunity hiding in plain sight. It is a shared entry point for Windows users, yet the person still has to connect the work between the apps.
Microsoft Global Hackathon 2026 / Participant
WHY THIS SURFACE
A launcher answers “Which app?” My question was different: could Start help someone understand what changed, see a useful next step, and choose how far the system should go?
TODAY / OPEN THE TOOLS
You connect the dots.
Teams
Outlook
Figma
PowerPoint
Open the review. Find the feedback. Rebuild context. Write the update.
THE CONCEPT / START THE WORK
A useful offer.
A visible choice.
“Draft ready · 1 decision needs you” brings related work together without quietly sending anything.
Prepare the next step. Keep the decision with the person.
A WORKED SCENARIO
Alex starts the day after an overnight US review. Maya needs an update before the 11:30 critique. Figma feedback adds a larger-text check—and an ambiguous request to make the layout denser.
Alex is a fictional scenario used to make the patterns concrete, not a research participant.
LAST NIGHT
Option B was selected in the US review.
THIS MORNING
Larger text is clear. “Denser” needs clarification.
11:30
Maya needs Alex’s update before the session.
DESIGNING FOR THE AGE OF AGENTS
During the hackathon, I ran a Friday learning session to help teammates apply agentic-experience patterns to their own submissions. Start became my worked example of the same method.
The “My task, redesigned” worksheet begins with a task—not an AI feature. It asks where help enters, what the person can inspect, and how they keep control.
Capability is not permission.
The seven-pattern exercise from my Friday learning session.
THE METHOD, IN THE INTERFACE
Each detail below comes from the actual Start prototype. Read the design decision on the left; inspect its expression on the right.
01 / ENTRY POINT
Meet the task.
Not another blank prompt.
The morning handoff card leads with a state: “Draft ready · 1 decision needs you.” It brings the review, the recipient and the next decision into one place.
Why: starting with relevant work makes the offer understandable before the person has to ask for help.
02 / PREVIEW
Show what will happen
before it happens.
The review makes the recipient, subject and message visible before sending. It also names the consequence: “Sends one email. Sending cannot be reliably undone.”
Why: a confident approval depends on seeing the action and its limits—not just a reassuring button.
03 / PERMISSION
Ability to act is not
permission to act.
The task has an explicit autonomy preference. “Draft for review” prepares local work; sending, sharing and changing access still require the person.
Why: the boundary belongs to this task. Connected work sources do not mean unrestricted access to someone’s personal life.
04 / EVIDENCE
Make the reason
inspectable.
“Why this?” names the Teams review and Figma feedback behind the offer. The person can open the review excerpt or comments instead of accepting an unexplained summary.
Why: provenance helps people judge a recommendation. Even a connected source can be incomplete.
05 / FORM
Give each decision
the right interface.
A quick composer shows the email fields that matter. Source inspection gets an evidence view. Permissions get their own controls. The experience is not forced into one endless chat.
Why: adapt the form to the work while keeping the action familiar and the current state visible.
06 / RECOVERY
Offer a way back.
Be honest about its reach.
Activity preserves the local brief and its earlier version. After an email is sent, the recovery path is a correction draft—not a promise that the message can be recalled.
Why: restoring local work and undoing an external action are different promises. The interface must say which one it can keep.
07 / ESCALATION
When the meaning is unclear,
ask one useful question.
“Make the layout denser” could mean tighter card spacing or more grid columns. The agent pauses and asks Alex to choose. No shared design or code changes while it waits.
Why: uncertainty should narrow the next decision, not become permission to guess.
NOW PUT IT TOGETHER
Explore the complete flow: inspect a morning handoff, review its evidence, set a boundary, resolve ambiguity and follow what happens after an action.
WHAT I TAKE FORWARD
The sprint turned a broad framework into decisions people could see and try. My contribution was both the prototype and a way for teammates to apply the patterns to their own work.
Can people distinguish a local draft from a sent action? Do they inspect evidence when it matters? Are the autonomy choices clear without explanation?
This is a one-week exploration, not a validated product outcome. Those questions are the next step—not results I can claim.
