Omar Motala

Builds AI products, public policy dashboards, legal analytics, and agent infrastructure.

Yale University · Harvard Law School

Six projects

01Agent coordination system

COORD

One SQLite-WAL database is the lifecycle authority; CLI, Python and MCP clients share it.
Leased claims, version-fenced handoffs, completion gated on declared artifacts, independent review.
Tracked CPU, GPU, API and MLX jobs; Board, Mesh, Map, Atlas, macOS menu bar and iOS views.

02Apartment comparison tool

ApartmentScout

11 weighted fit dimensions; rent adjusted for concessions and fees.
Door-to-door commute by time of day; NYC building records and complaint patterns.
Modeled sun by floor, date and time; floor plans and furniture in synchronized 2D and 3D.

03Native iOS app

Telly

Taste profile and viewer archetype from watch history; rates titles; recommends what to watch.
Ask Telly and Concierge — natural-language search and recommendations.
SwiftUI iPhone app; AWS Lambda, DynamoDB, Cognito, API Gateway, EventBridge.

04Public policy dashboard

Climate Hostility

US state climate and ESG legislation; 1,524 annual bill attempts surfaced.
Filters — anti-ESG / pro-ESG; three policy universes; binding, symbolic, pending, dead.
Views — net-hostility map, state rankings, trends, topic flows, state profiles, exports.

05Travel planner

Trippy

City order, nights and gateways compared across air, rail and car legs.
Cash versus points — fares, awards, taxes, transfer ratios and valuations together.
Open source, TypeScript and Next.js; exports to HTML, PDF, JSON, CSV.

View project ↓
Source ↗

SCOTUS OT2025 per-justice scorecard header: 148 of 171 justice votes and 17 of 19 case outcomes right, above the first rows of the case grid.

06Legal prediction model

SCOTUS Prediction

Two stages — each justice’s affirm/reverse vote, then the case outcome.
OT2025 — 148 / 171 justice votes correct (87%).
17 / 19 case outcomes correct; scope is those 19 decided cases.

View project ↓

View full size

01Agent coordination system

COORD

Local coordination system for coding agents (Claude, Codex) and tracked jobs. One SQLite-WAL database is the lifecycle authority. Public source, MIT.

2Provider usageCapture loop · 18 sThe usage capture, scrolled. Dollars are recorded token-cost estimates, not money paid. Full capture Original PNGClaude Personal, Claude Team and Codex quotas; session and weekly resets; runout estimates; daily token-cost history.

Uncropped COORD macOS menu-bar dropdown: CPU, GPU, RAM and disk readouts, Claude and Codex usage bars, active work by agent and the Next Up queue. — **3macOS menu bar**Uncropped capture; scroll inside the frame. Original PNGCPU, GPU and RAM; active work by agent; the queue. Native views are read-only by default.

Flattened 2D COORD mesh: a ring of agent and work nodes with fan-out spokes and labelled module clusters. — **4Mesh**Live projection, flattened to 2D: agents, work, dependencies and evidence.Live read-only projection of agents, work, dependencies and evidence, grouped into labelled modules.

Authority

One SQLite-WAL database is the lifecycle authority.
CLI, Python and MCP clients share it.
Public source, MIT.

Handoffs and review

Leased claims; version-fenced handoffs.
Completion gated on declared artifacts.
Independent review — no self-approval.

Jobs

Tracked CPU, GPU, API and MLX jobs.
Process-derived liveness; resource locks.
Dead-run recovery.

Views and usage

Board, Mesh, Map, Atlas, macOS menu bar, iOS over the same state.
Native views read-only by default.
Token-cost history — recorded estimates, not money paid.

Captures: mesh, usage and menu views come from the author’s installed environment. The public repo ships the generic system with synthetic demo data.

Originals, uncropped · select to enlarge

COORD mesh: hundreds of red and blue work nodes, gold evidence diamonds and labelled module clusters on a black grid. — Mesh original · PNG

Uncropped COORD provider usage capture: Claude and Codex quotas, resets, runout estimates and daily token-cost estimates. — Provider usage · PNG

COORD native usage window: total token-cost estimate, Claude and Codex session and weekly quota bars and orange and violet daily cost-estimate histograms. — Native usage window · recorded token-cost estimates

COORD Map in Orbit lens: a luminous radial sunburst of child work items around one epic. — Map, Orbit lens

COORD iOS home with running, attention, next and done counts and the work running now. — iOS

Native macOS COORD Cockpit window listing running, attention and planned work with progress bars and priorities. — macOS Cockpit · read-only by default

COORD Operations Atlas with a read-model rail, summary strip and live topology, demo data. — Ops Atlas · demo data

COORD mesh beside the provider-usage tracker and the menu-bar widget on a tinted stage. — Mesh, usage and menu

COORD macOS Cockpit board listing running, attention and planned work, beside the menu-bar widget and the provider-usage tracker. — Board, usage and menu widget

COORD critical-flow mesh: a red hub with edges fanning into dependency layers. — Mesh, critical flow

Full COORD mesh workspace with cluster roster, graph and reader panel. — Mesh workspace

COORD swarm mesh grouped by owner, from the public demo with synthetic data. — Mesh by owner · demo data

COORD usage cost history: token totals and a daily cost-estimate chart. — Cost-estimate history

COORD comms view: recorded traffic direction graph and event feed. — Comms traffic

COORD Board of tracked jobs, from the public demo with synthetic data. — Board, jobs · demo data

Details

Lifecycle authority — one SQLite-WAL database; CLI, Python and MCP clients share it.
Claims — leased claims and version-fenced handoffs; completion gated on declared artifacts.
Review — independent; no self-approval.
Jobs — CPU, GPU, API and MLX; process-derived liveness, resource locks, dead-run recovery.
Usage — Claude Personal, Claude Team and Codex quotas; session and weekly resets; runout estimates; daily token-cost history (estimates).

Spec

Type	Agent coordination system
Views	Board, Mesh, Map, Atlas, macOS menu bar, iOS
Source	COORD-Harness ↗ · MIT
Captures	Author’s environment; public repo uses synthetic demo data

View full size

02Apartment comparison tool

ApartmentScout

Renter’s tool for comparing an apartment shortlist. Ranks listings on 11 weighted fit dimensions, with commute, building records and modeled sun.

2Unit in 3DLoop · 12 sThe furnished living room, panning toward the bay window and the city. Modeled interior.Floor-plan tracing, furniture placement and clearance checks, in synchronized 2D and 3D.

Side-by-side modeled window views from the same building: floor 8 in shadow and floor 25 in direct sun. — **3Sun by floor**Same window, same moment: floor 8 in shadow, floor 25 in direct sun.Modeled sun and shadow by floor, date and time, from building footprints and heights.

Modeled 3D Midtown skyline at dusk with landmark labels and the unit’s tower highlighted in blue. — **43D city at dusk**Modeled Midtown at dusk; landmarks labelled, the unit’s tower in blue.Modeled city around the unit: its tower and window marked, landmarks labelled, day and night orbits.

5Walkthrough72 seconds, with controls and sound. Try the interactive demo ↗The 72-second walkthrough; the interactive demo runs in the browser.

Rank

11 weighted fit dimensions; unknowns stay visible.
Rent adjusted for concessions and fees.

Location

Door-to-door commute by time of day, to multiple destinations.
Building records and complaint patterns (NYC).

Sun and interiors

Sun and shadow by floor, date and time.
Tracing, furniture, clearances; 2D and 3D in sync.

Decide

Up to 4 finalists; leasing-office questions.
Decision memo; phone tour-day plan.

Example: a fall-2026 Manhattan shortlist. Interiors, sunlight and commutes are models and estimates; prices are historical examples.

Loops

Ranked list, heat map, shortlist · 24 s

Unit in 3D, toward the window · 24 s

Sun by floor, floor by floor · 10 s

Sun and city · select to enlarge

Heat table of direct sun hours on the 15th of each month for every floor of the building, with the unit’s floor highlighted. — Sun hours by floor and month

Two modeled 3D views of the same block at noon on June 21 and December 21, showing longer winter shadows. — Summer vs winter, noon shadows

3D city orbit by day: the unit’s tower highlighted, a sight line to the window on floor 25 and a direct-sun readout. — City orbit, day

The same modeled 3D city orbit at night with lit windows and the candidate tower highlighted. — City orbit, night

Modeled view from the unit’s window: towers, a park canopy and the river, with a direct-sun chip. — View from the window, floor 25

3D city building models in blue with the unit’s tower highlighted, a sightline to the window on floor 25 and a sun-angle readout. — Building models, sightline

Interiors, map and decision · select to enlarge

ApartmentScout dashboard with ranked Manhattan listings and a commute heat map, with a large card of the furnished 3D interior of the open listing. — Dashboard and interior

Modeled furnished living room with a dining table, sofa, desk and bay windows onto the city. — Living room

Modeled bedroom seen from the door with corner windows onto a park and river. — Bedroom

Modeled living room toward the desk and bay window with sun streaks. — Desk and bay window

Modeled night view from the window over lit towers. — Window at night

Unit 3D inside the app with its controls and time-of-day bar. — Unit 3D in the app

Ranked listings, commute heat map over Manhattan and a listing drawer with monthly cost. — Heat map and listing drawer

Composite of the heat-map dashboard, the 3D city orbit and the furnished interior. — Map, city and interior

Dashboard map in 3D city mode. — Map in 3D

Modeled sun reaching a high-floor window. — Sun on a high floor

Plan view of modeled sun and shadow around the building. — Sun, plan view

Sun path chart: seasonal arcs over the surrounding skyline silhouette. — Sun path chart

Floor-plan studio: traced plan with furniture placed beside a 3D preview. — Floor-plan studio

Floor-plan studio with modeled sunlight across the unit. — Studio sunlight

Two finalists’ modeled window views compared side by side. — Side by side in 3D

Comparison table of four finalists in lettered columns with photos, rents, fit scores and feature rows, best-in-row values marked. — Four finalists

Building research page with public records and complaint patterns. — Building records

Map with walking-distance rings around a listing. — Walk rings

Tour-day planner timeline with travel between homes. — Tour planner

Printable decision memo summarising the ranked shortlist. — Decision memo

Phone tour-day plan with the day’s homes, times and a checklist. — Phone tour day

Details

Rank — 11 weighted fit dimensions; rent adjusted for concessions and fees; unknowns stay visible.
Commute — door-to-door by time of day to multiple destinations.
Buildings — records and complaint patterns (NYC).
Interiors — floor-plan tracing, furniture placement and clearance checks in synchronized 2D and 3D.
Sun — modeled sun and shadow by floor, date and time from building footprints and heights.
Decide — up to 4 finalists; leasing-office questions; printable memo; phone tour-day plan.

Spec

Type	Apartment comparison tool
Demo	Interactive demo ↗
Source	github.com/0marm0/ApartmentScout ↗
Media	72 s walkthrough; 4 silent loops; 30 screens

View full size

03Native iOS app

Telly

tellytech.ai ↗

SwiftUI iPhone app in TestFlight beta. Builds a taste profile and viewer archetype from watch history, rates titles and recommends what to watch.

Product demo51 seconds, with sound controls.

Fight Club title screen: where-to-watch row and a 3 by 3 grid of provider scores labelled adjusted for you. — **Provider ratings**Fight Club title screen: where-to-watch row and the 3×3 multi-provider ratings grid (IMDb, Letterboxd, Trakt, MDBList, TMDB, RT Critics, Metacritic, RT Audience, Ebert), adjusted for you.

tellytech.aiThe beta site.

Taste

Taste profile and viewer archetype from watch history.
Rates titles; recommends what to watch.

For You home feed with a rated poster rail and a list of recommendations. — **On iPhone**Title screens with scores from several rating providers, adjusted for you, and three silent loops from the demo.

Captain America: Civil War title screen with poster, provider scores, a Telly score and an Add button. — **On iPhone**Title screens with scores from several rating providers, adjusted for you, and three silent loops from the demo.

Ask Telly and Concierge

Natural-language search.
Natural-language recommendations.

App · Backend

SwiftUI, iPhone; TestFlight beta.
AWS — Lambda, DynamoDB, Cognito, API Gateway, EventBridge.

Gallery · select to enlarge

Interstellar sheet with streaming providers and a grid of provider scores adjusted for you. — Adjusted ratings

Viewer archetype screen with watch time, decision signals and genre breakdown. — Viewer archetype

Concierge sheet with type, length and mood chips and a natural-language prompt. — Concierge

Rate Faster screen with twenty selected titles and a Rate button. — Rate faster

Rating training poster grid with movie and TV tabs and provider filters. — Rating training

Library grid of posters with ranked rating badges. — Library

Severance title screen with the rating slider set to 10. — Rating a title

Supergirl sheet with raw and adjusted rating grids. — Raw vs adjusted

Site on mobile

Three Telly iPhone screens fanned on gold light. — Three-phone composite

Loop

Ranked library · 4 s

Details

Taste profile and viewer archetype built from watch history.
Ratings for titles; recommendations for what to watch.
Ask Telly and Concierge — natural-language search and recommendations.
Backend — AWS Lambda, DynamoDB, Cognito, API Gateway, EventBridge.

Spec

Type	Native iOS app
Client	SwiftUI, iPhone; TestFlight beta
Site	tellytech.ai ↗
Media	51 s demo; 4 silent loops; 15 screens

View full size

04Public policy dashboard

Climate Hostility

Public dashboard of US state climate and ESG legislation. 1,524 annual bill attempts surfaced.

Live dashboardLoop · 23 sMap toggles anti / pro / net; rankings and annual trends follow.

Filters

Direction — anti-ESG or pro-ESG.
Policy universe — Corporate/Fiduciary ESG, Climate, Broad.
Status — binding, symbolic, pending, dead.

Views

National net-hostility map; state rankings at top 10 / 15 / 25.
Counts — combined, enacted, attempts.
Annual trends by status and mechanism; topic flows.

Topics

19 distinct topics; 7,910 clause assignments.
Anti-leaning 12 topics; pro-leaning 6.
Keyword trends — top 10 ESG keywords by bill hits, by year.
Topic evolution — annual clause assignments by topic.
Topic × stance — per-topic anti / pro clause split.

State profileTexas: composition, focus, mechanisms.

Bill drawer for Texas SB13: operative summary, status and stance tiles, mechanism and reach flags, severity bars and source passages. — **Bill detail**TX SB13. Scroll inside the frame.

Drill-down · Stack

State profiles, representative bills, exports.
Data modeling, policy taxonomy, static deployment.
Companion paper, Harvard Law.

Loops

Drill-down

State profile, scrolled · 10 s

Mechanisms view, scrolled · 9 s

Gallery · select to enlarge

Climate Hostility national landscape: summary count tiles, a US map shaded from pro-ESG blue to anti-ESG red, and a ranked bar chart of the top states. — National landscape

Stance and net view: annual anti-ESG and pro-ESG bar chart with enacted lines, summary tiles and a state-by-year activity heatmap. — Stance and net

Mechanisms view: diverging anti versus pro bar chart by primary mechanism above a stacked area chart of mechanism shift over time. — Mechanisms by stance

Climate Hostility overview page with navigation, summary tiles, the national map and the state ranking. — Overview

Dashboard and state profile

All-state status trend by year: binding, attempt and dead lines for anti and pro bills. — Status trend by year

Line chart of enacted bills by year, anti in red and pro in blue. — Enacted bills by year

Filters and counting drawer with universe, status, focus-area and stance chips. — Filters and counting drawer

Expanded header panel with four key-findings cards and four methods-at-a-glance cards describing counting grains, classification and models. — Key findings and methods

Bar chart of top states by bills in the current slice, led by Texas, New York, Missouri and Oklahoma; red bars anti-leading, blue pro-leading. — Top states

Topic evolution chart across years. — Topic evolution

Explore view with a mechanism by focus grid of passage rates. — Explore

Framework view with tier filters and tier by pathway and tier by mechanism heatmaps. — Policy taxonomy framework

Texas profile, full · scroll

Details

Coverage — US state climate and ESG legislation; 1,524 annual bill attempts surfaced.
Filters — anti-ESG / pro-ESG; Corporate/Fiduciary ESG, Climate, Broad; binding, symbolic, pending, dead.
Views — net-hostility map; top 10 / 15 / 25; combined, enacted and attempt counts; status and mechanism trends; topic flows.
Drill-down — state profiles; representative bills; exports.
Paper — State ESG Legislation and Financial Judgment, Harvard Law.

Spec

Type	Public policy dashboard
Stack	Data modeling, policy taxonomy, static deployment
Live	esg-hostility.vercel.app/v2 ↗
Media	4 silent loops; 17 screens

View full size

05Travel planner

Trippy

Open-source multi-city travel planner (TypeScript, Next.js; beta). Compares city order, nights and gateways across air, rail and car legs.

Explore each city: seven stops with idea counts and nights, a strip of activity photos and the first stop’s city card above its waypoint map. — **2Cities**Seven stops with idea counts and nights, a photo strip, and the city card above its waypoint map.Each stop with its nights, idea count, photos and a waypoint map.

Cash versus points workbench for Kraków to Brussels: direct and connecting cash fares, award points, taxes and positioning, giving best cash, award cash cost and cents per point. Sample inputs. — **3Cash vs pointsSample inputs**Kraków to Brussels: best cash, award cash cost and cents per point.Fares, awards, taxes, transfer ratios and point valuations recorded together; cents per point. Sample inputs.

Kraków activities: category chips, must, should and could filters and photo cards with time allowances and Add to day. — **4Activities**Kraków: category chips, must / should / could filters, photo cards with time allowances.Must, should and could tiers; photo cards added to days.

Export studio previewing a visual itinerary for the 2026 Europe trip with a large aerial photo. — **5Export**Export studio: visual dossier or compact brief, with a photo cover.HTML, PDF, JSON and CSV. Six booking states kept distinct: a quote never becomes a confirmation.

How the optimizer works

Ranks city orders. Tries every order of the middle cities, first and last fixed (up to 9 stops); takes the best recorded option per leg; scores each route 1–99 with confidence and known cash.
Weighted trade-off. Cost, time, transfers, reliability and carbon, weights summing to 1; Balanced value, Lowest cost and Shortest practical time presets.
Gaps count against a route. Each missing leg costs 11 points; backtracking is penalised and stated.
Cash vs points. Points spent from transfer ratio, minimum and block size; cents per point = (best cash − award taxes − positioning) ÷ points; award favoured at 1.5¢+.
Multi-leg scenarios. Gateways incl. open jaws; through, separate or mixed tickets; cash per currency and points per programme.
Must / should / could. Activities tiered, dragged onto days, grouped within 1.5 km.

Itinerary, stays, output

Dated itinerary editor with photo activity cards.
Hotel stays with price basis and cancellation terms.
Shared board — ideas, votes, reviewed changes.
Scenarios, undo/redo, snapshot restore.
Exports — HTML, PDF, JSON, CSV.

Note: screens show recorded sample plans, not live inventory or bookings. The hosted demo is older and may lag the source.

Loops

City maps

Optimizer and points · sample inputs

City waypoint maps · select to enlarge

Kraków city page: hero photo and a map with the airport, Kraków Główny station, the apartment and sight pins. — Kraków · waypoints

Berlin city page: map with the Hauptbahnhof and sight pins across Mitte and Tiergarten. — Berlin · waypoints

Wrocław city page: Old Town map with the hotel, Wrocław Główny station and sight pins. — Wrocław · waypoints

Amsterdam city page: canal-belt map with Centraal station and sight pins. — Amsterdam · waypoints

Munich city page: central map with the Hauptbahnhof and sight pins. — Munich · waypoints

Nürburgring stop: aerial track photo and a map with the hotel and activity pins. — Nürburgring · waypoints

Optimizer, stays and output · select to enlarge

Trip calendar for September and October 2026: each day shows its city, with icons for logistics and the number of sights. — Trip calendar · recorded sample plan

Itinerary editor for a Kraków day: a confirmed stay, a train needing booking and photo activity cards, beside ideas grouped must do and should do. — Itinerary editor

Trippy route map of a 2026 Europe trip with numbered stops at Amsterdam, Brussels, Nürburgring, Frankfurt Airport, Munich, Berlin, Wrocław and Kraków. — Route map, full workspace

Whole-trip route map with numbered stops and flight, rail and drive legs. — Route map

Route map beside the itinerary editor and the export studio on a tinted stage. — Route, itinerary, export

Decision lab: leg, objective, modes and three ranked city orders. — Decision lab

Transport options from Kraków to Brussels. — Transport by leg

Alternatives listed for each leg of the route. — Alternatives by leg

Map of flight paths between gateways. — Flight paths

Three hotel stays compared with price ranges, basis and cancellation terms. — Hotel stays compared

Activity detail drawer with photo, time allowance and day assignment. — Activity detail

Day map for Amsterdam with the day’s stops. — Day map

Day geography grouping a day’s stops by area. — Day geography

Decide mode of the trip workspace. — Decide mode

Collaboration editor with suggested places and proposed days. — Shared board

Trip preparation: tasks, packing and plan B options with photos. — Preparation

Optimizer details

Like with like. Two quotes are ranked only if currency, dates, route, party size, rooms, cabin, baggage and cancellation terms match; otherwise the reasons are listed.
Unknowns stay unknown. Missing cash, points or valuation inputs are not inferred.
Scenario checks. Flags unknown ticketing, missing gateways and legs out of order.
Proposals for your trip. Up to 6 city orders (original, reversed, rotated), each with balanced, longer-first or longer-last night splits within your min and max nights.
Day order. A nearest-neighbour order from the hotel is kept only when shorter than the recorded one; unpinned activities are listed separately.
Limits. Results are planning comparisons, not live fares or a proven global cheapest route.

Spec

Type	Open-source travel planner; beta
Stack	TypeScript, Next.js
Legs	Air, rail, car; elapsed time, transfers, coverage gaps
Output	HTML, PDF, JSON, CSV; 6 booking states
Source	Trippy-Travel-Planner ↗

SCOTUS OT2025 per-justice scorecard: title, tiles reading 148 of 171 justice votes right and 17 of 19 case outcomes right, a legend and the first case rows under nine justice headshots.

View full size

06Legal prediction model

SCOTUS Prediction

Predicts each justice’s affirm/reverse vote, then the case outcome. OT2025: 148 / 171 votes and 17 / 19 outcomes correct.

SCOTUS OT2025 per-justice scorecard: a grid of 19 decided cases by nine justices, each predicted vote green when right and red when wrong, with a right-count column. — **Per-justice scorecard**All 19 decided cases; green right, red wrong.

Model

Stage 1 — each justice’s affirm/reverse vote.
Stage 2 — the case outcome.

OT2025 scorecard

Justice votes — 148 / 171 correct (87%).
Case outcomes — 17 / 19 correct.
Scope — only those 19 decided cases.

Scorecard stat strip: 148 of 171 justice votes right, 87 percent across 19 cases; 17 of 19 case outcomes right, missed Chatrie and Watson. — **Stat strip**From the scorecard header.

Scorecard with best-model confidence: every justice cell shows the predicted call and a confidence percentage; misses shaded red. — **With model confidence**Each call with its confidence; misses in red.

Heatmap of nine justices across six doctrinal axes with signed positions, blue liberal-leaning and red conservative-leaning; a model input. — **Ideology axes**Justice × six doctrinal axes. A model input.

Gallery · select to enlarge

Full dark scorecard with filled, confidence and detail tables for the 19 cases. — Dark scorecard, full · scroll

Light-theme per-justice scorecard for the 19 cases. — Scorecard, light

Light-theme confidence scorecard for the 19 cases. — Confidence, light

Parallel-coordinates chart of the nine seats across six doctrinal axes. — Seats across axes · model input

Details

Two stages — each justice’s affirm/reverse vote, then the case outcome.
Votes — 148 / 171 correct (87%).
Outcomes — 17 / 19 correct.
Scope — only the 19 decided OT2025 cases on the scorecard.

Spec

Type	Legal prediction model
Term	OT2025, 19 decided cases
Inputs shown	Ideology axes heatmap; seats across axes
Link	No public link

Scholarship

Three papers · DOCX

State ESG Legislation and Financial Judgment

Harvard Law research paper; companion to the Climate Hostility dashboard.

Download DOCX ↓

ESG Regulatory Divergence: EU, SEC, and California Climate Disclosure

Harvard Law research paper.

Download DOCX ↓

Redefining Corporate Purpose

Yale thesis.

Download DOCX ↓

About and contact

omar-mo.com

Omar Motala builds AI products, public policy dashboards, legal analytics, and agent infrastructure.Yale University · Harvard Law School

Spec sheet

Six projects · index and capabilities

Project index
Project	Type	What it does	Key figure	Stack	Media	Link
01COORD	Agent coordination system	Coordinates coding agents (Claude, Codex) and tracked jobs through one SQLite-WAL database.	6 views over one database	SQLite-WAL; CLI, Python, MCP; macOS, iOS	1 capture loop; 19 screens; 3 original PNGs	Source ↗
02ApartmentScout	Apartment comparison tool	Ranks a rental shortlist on weighted fit, effective rent, commute, building records and modeled sun.	11 weighted fit dimensions	Interactive web demo; synchronized 2D and 3D	72 s walkthrough; 4 silent loops; 30 screens	Demo ↗Source ↗
03Telly	Native iOS app	Builds a taste profile and viewer archetype from watch history, rates titles and recommends what to watch.	TestFlight beta	SwiftUI; AWS Lambda, DynamoDB, Cognito, API Gateway, EventBridge	51 s demo; 4 silent loops; 15 screens	tellytech.ai ↗
04Climate Hostility	Public policy dashboard	Maps and ranks US state climate and ESG legislation by direction, policy universe and status.	1,524 annual bill attempts surfaced	Data modeling, policy taxonomy, static deployment	4 silent loops; 17 screens	Live ↗
05Trippy	Travel planner	Ranks whole-route city orders, compares cash versus points and hotel stays; shared board for group planning.	6 booking states kept distinct	TypeScript, Next.js; open source; beta	2 silent loops; 28 screens	Source ↗Old demo ↗
06SCOTUS Prediction	Legal prediction model	Predicts each justice’s affirm/reverse vote, then the case outcome.	148 / 171 votes (87%); 17 / 19 outcomes	Two-stage prediction: votes, then outcome	9 screens	No public link

01COORD

Authority	One SQLite-WAL database; CLI, Python, MCP
Claims	Leased claims; version-fenced handoffs
Completion	Gated on declared artifacts; independent review
Jobs	CPU, GPU, API, MLX; liveness; locks; dead-run recovery
Views	Board, Mesh, Map, Atlas, macOS menu bar, iOS
Usage	Quotas, resets, runout and token-cost estimates

02ApartmentScout

Ranking	11 weighted fit dimensions; unknowns visible
Rent	Adjusted for concessions and fees
Commute	Door-to-door, by time of day, multiple destinations
Buildings	Records and complaint patterns (NYC)
Interiors	Tracing, furniture, clearances; 2D and 3D
Sun	By floor, date and time from footprints and heights
Decide	Up to 4 finalists; questions; memo; tour-day plan

03Telly

Taste	Profile and archetype from watch history
Ratings	Rates titles
Recommend	What to watch next
Language	Ask Telly and Concierge: natural-language search and recommendations
Client	SwiftUI, iPhone; TestFlight beta
Backend	Lambda, DynamoDB, Cognito, API Gateway, EventBridge

04Climate Hostility

Coverage	US state climate and ESG bills; 1,524 annual attempts
Direction	Anti-ESG / pro-ESG
Universes	Corporate/Fiduciary ESG; Climate; Broad
Status	Binding, symbolic, pending, dead
Views	Net-hostility map; top 10/15/25; counts; trends; topic flows
Drill-down	State profiles; representative bills; exports

05Trippy

Routing	City order, nights, gateways; air, rail, car
Optimizer	Whole-route ranking; weighted cost, time, transfers, reliability, carbon
Points	Fares, awards, taxes, transfer ratios, valuations; cents per point
Stays	Price basis and cancellation terms
Group	Shared board; scenarios; undo/redo; snapshots
Output	HTML, PDF, JSON, CSV; 6 booking states

06SCOTUS Prediction

Stage 1	Each justice’s affirm/reverse vote
Stage 2	Case outcome
Votes	148 / 171 correct (87%)
Outcomes	17 / 19 correct
Scope	Only the 19 decided OT2025 cases