Choose the QR experienceThe preview and downloaded QR update immediately.
App Clip QR selected · Opens the iPhone App Clip directly.
Share this link or QR at check-in. On iPhone it can open the Mixer App Clip for this event.
Admission
Registration is open
Guests can join and update their approved event profiles.
CreatedRegistrationClosed
Matching reveal
Newcomers enter immediately
Queued
Mixer scores each approved newcomer, reranks affected attendees, and retries unfinished work every 15 minutes. A full refresh remains available as an organizer fallback.
Preparing comparisons…0%
At least two approved profiles are required.
Event experience
Program
Loading program…
This changes the event experience, not existing attendee answers or their privacy settings.
Live event Q&A
Audience questions
Questions appear anonymously and are ranked by audience votes. Enable a moderator link to manage them without a Mixer login; attendees see answered questions in the App Clip or Mixer app. Hide removes a question from the audience without deleting it.
?
No questions yet
Audience questions will appear here during the event.
Balanced introductions
Match lookup
Select an attendee to see their introductions, match status, fit score, and public shared context without exposing private profile answers.
↔
Choose an attendee
Select someone above to inspect their introductions.
Private organizer tools
Synthetic demo data
Your account only
Create a convincing event room for product demos, including category analytics. These controls are only available in your web organizer account.
Loading synthetic data status…
Development demo
Demo setup
Not prepared
Create an isolated Development mirror, make it the Debug App Clip event, and generate ten event-aware profiles. Synthetic attendees never enter Production.
0 of 10 profiles ready0%
0Demo attendees
0Approved profiles
Ready to mirror this event and generate its demo room.
Safe to press again after editing the event; Mixer updates the same ten synthetic slots.
↗
Chat demonstration
After matching, the best generated attendee will send an incoming connection request. Accept it in the App Clip, then send your one-way message in the real chat.
Event setup
Profile configuration
0Profile fields
0Custom questions
0Meetup zones
No event goals added yet.
Event room
Directory
Select an attendee to see their approved answers to every event question. Owner-only answers stay out of the Room and shared directory.
♟
No attendees yet
Share the invitation link or QR to start building the room.
Organizer-defined categories assigned privately from each approved AI profile.
0profiles
How they connected
Connection pathways
Aggregate
See which attendee groups were introduced, connected, and started conversations.
Connected groups
◎
Add attendee categories to unlock audience intelligence
Create neutral labels such as Founder, Investor, Mentor, or Job seeker in Event settings. Mixer’s AI suggests one and shows it for approval without adding another App Clip question.
Shared signals
Top overlap topics
Aggregated
Recurring themes across approved profiles that appear in strong matches.
Use reciprocal profile fit and balance introductions across the room.
Mixer treats this as pairing guidance only and never uses protected traits or reveals private answers.
Profile fields
General fields such as Company Name and Title are collected in the approved profile. Email, LinkedIn, Website, Instagram, X, and TikTok toggles add validated contact actions; they do not appear under “Find your people.”
Organizer questions & Room mapping
Public toggles control whether an approved answer appears on the Room card, full profile, or category filter. Private answers stay out of both the Room and shared directory; all approved event-question answers remain available to the event owner.
After the event
Response categories
Organizer only
Run AI bucketing after the event closes, review every original answer, and correct any assignment. These buckets never replace what attendees wrote on their cards.
Room profile categories
The AI suggests one event-relevant label during profile return. After attendee approval, it appears as a badge at the top of their Room profile.
Named meetup zones
Named public areas help attendees meet without collecting precise location.
Event artwork
After the event
Shareable directory page
Not created
Publish a read-only attendee-facing page after the event. It never links back to this dashboard and includes only approved profiles and fields you marked public.
Standard event directory links remain available for 30 days after the event date.
Archive event
The invitation stops working and this event disappears from your organizer list. This cannot be undone in the portal.
This address never changes. When you set a different event live, the QR immediately opens that event.
Use this URL on signs, badges, and printed flyers. Do not replace it with an individual event link.
Live destination
Current event
Live
M
EventDate not set
Scans of the permanent QR currently enter this event.
Occurrences
Events in this series
Each occurrence keeps its own attendees, matches, conversations, and analytics. Create ahead of time, then choose Set live when you are ready.
Development onlyMATCHING TEST LAB
Test matching before it reaches an event.
Test individual profile pairs, compare models and candidate selection, then validate the chosen setup with a complete production event.
Isolated model test
Compare the same attendee pairs
Real OpenAI calls
Each selected model receives the same profiles and prompt, so differences in speed, cost, and match quality are directly comparable.
Live run
Ready
Idle
Choose models and start an isolated comparison.
Preparing comparisons…0%
Models
—
Comparisons complete
—
Failures
—
Results update automatically as each model finishes.
Model results
Speed, quality, and OpenAI cost
No run selected
Expected matches found compares each model’s returned matches with the hidden synthetic answer key. Average response is one pair’s wall time after work starts; queue wait is separate, and concurrency changes throughput rather than that individual response.
◎
No model results yet
Run a comparison to see one clear result card per model.
Final stage · Selection Sandbox
Build the introduction roster
Saved outputs only
Rebuild the final roster from every saved pair score in this run. Changing these controls never calls OpenAI or adds cost.
Model output similarity—
Agreement, not match quality
Same recommendations—40% of similarityRanking agreement—35% of similarityScore agreement—25% of similarity
See exactly who each model recommends for one person and how shared recommendation scores move.
— similarity0 shared0 only Model A0 only Model B
Model A0 recommendations
◎
No recommendations
No qualifying introductions under these rules.
Model B0 recommendations
◎
No recommendations
No qualifying introductions under these rules.
Run history
Model comparisons
0 runs
Select a run to inspect its model results and sample outputs.
Started
Status
Models
Attendees
Matches
Output
Concurrency
◷
No comparisons yet
Completed and active Model Lab runs appear here.
Step 1 · Profile generator
Create controlled test attendees
Ready
Describe the roles and relationships you want to test. Mixer creates editable fictional profiles and keeps them in your reusable library.
Step 2 · Pair scorer
Score any two profiles
Ready
Choose two saved profiles and a model. The model receives the same production scoring prompt used by Model Lab.
Latest comparison
Pair score and reasoning
No result
◎
No pair scored yet
Select two profiles and run the exact production pair-scoring prompt.
Match score——
Response time
—
Total tokens
—
API cost
—
Points in common
Complementary strengths and needs
Reusable profile library
Generated test attendees
0 profiles
Click a profile to inspect it, edit any field, or assign it directly to side A or B.
◇
No test profiles yet
Generate a controlled group above to start comparing pairs.
Saved experiments
Comparison history
0 comparisons
Every result keeps the exact profile snapshots, model, score, reasoning, timing, tokens, and cost from that run.
Profiles
Model
Score
Response
Tokens
Cost
Status
Created
◷
No saved comparisons
Completed and active pair tests appear here.
Candidate selection test
Find the best people before matching
Astra benchmark
Every selected method returns the same number of candidates. Astra scores every possible pair once, creating the reference used to measure which methods keep the best matches.
Live run
Ready
Idle
Choose candidate methods and confirm the Astra benchmark.
Preparing benchmark…0%
Methods
—
Work complete
—
Failures
—
Results update automatically as each selection method finishes.
Candidate results
Quality, time, and cost
No run selected
Astra top matches recovered shows how many of the benchmark’s best matches survived each method’s candidate cutoff.
◎
No candidate results yet
Run a comparison to see one result card per method or selected lower-cost model.
Full event setup
End-to-end sandbox
Current production preset
Combine a Candidate Lab setup, pair-scoring model, roster rules, arrival rate, and real backend concurrency. Every selected operation runs; nothing is sampled or extrapolated.
Live run
Ready
Idle
Configure the event and start a complete production simulation.
Preparing event…0%
Attendees arrived
—
Attendees fully matched
—
Model comparisons complete
—
Failures
—
Results update automatically while the production workload runs.
Total event time—First arrival until matching work finishesFirst match—Average / P95 after arrivalFull roster—Average / P95 to reach the targetRoster update—Average / P95 after a new option arrivesTotal API cost—Candidate selection + pair scoringCandidate stage cost—Selector or embedding callsReached match target—Completed attendees / total attendees
Workload breakdown
Cost and time by stage
Measured calls
Candidate selection and pair scoring report their own requests, retries, tokens, response time, and cost. Roster selection replays saved scores locally.
Stage
Method
Requests
Average response
Failures
Retries
Tokens
Cost
◎
No stage results yet
Run a full event to separate candidate, scoring, and roster work.
Attendee experience
Match timing by event interval
No timeline
Each row shows what active attendees experienced during that interval: first matches, reaching their initial target, and meaningful roster updates as the room grew.
Event time
Active
First match avg / P95
Initial roster avg / P95
Update avg / P95
Updated
Waiting for first
Waiting for roster
◷
No attendee timeline yet
The timeline appears as attendees arrive and matches complete.
Run history
Production simulations
0 runs
Select a run to view its measured timing and token usage above.
Scoring model
Candidates
Concurrency
Selection
Min score
Status
Attendees
Arrival rate
Match target
Total time
Average
P95
Tokens
Cost
Completion
◷
No simulations yet
Run a full production event to measure its time and token usage.
