A working example · Upside Explorer

Find more upside in a deal, together.

A live model the whole team can work at the same time. Move an assumption and the numbers move with you, with your partners and a team of focused agents on the same model. Every number traces back to the inputs behind it.

Try the live demo → Book a call
Your work email opens the demo. It runs in your browser.  ·  Meet the team
The Upside Explorer on an invented sample deal: levers on the left, scenarios and a live dashboard in the middle, agents on the right
The Upside Explorer as it runs in the app today, on an invented sample deal. On the left, the levers: every assumption, grouped the way a deal is argued. In the middle, ten scenarios and the live model, recomputing as anything moves. On the right, the agents that propose moves and the Anchor that checks them. Each column is explained below ↓
What's on the screen

One deal, three columns, all on the same live model.

Left · the levers

67 assumptions you can move

Grouped the way a deal is argued: program, cost, revenue, operations, capital stack, and risk and returns. Move one and every number recomputes.

A dot shows who moved each lever, you or the local or live agents, and a tick marks where it started, so every change from the baseline stays visible.

Middle · the live model

Ten scenarios and the numbers a lender asks for

From an honest floor and the base case as underwritten, through a downside stress, to a ladder of indicative upside moves, each one applied to this deal's own baseline.

The dashboard shows levered IRR against the target, coverage, equity multiple and yield on cost, then the untrended spread, the number that survives diligence. Below it sit the capital stack and a sensitivity grid, and any version can be saved and printed as a report.

Right · the agents

A fleet that proposes; the engine decides

Six composers each look for moves in their own area. Run them locally, instantly and free, or live, with Claude reasoning each move and showing its cost.

Live moves wait for your review: the Anchor checks each one against the feasibility baseline and flags any that drift past what the study supports, and nothing is applied until you choose. Every number still comes from the fixed engine.

Four kinds of deal

One explorer, shaped to the kind of deal.

Each kind of deal gets its own levers, dashboard and agents on the same live, tested engine, so a conversion is argued like a conversion and a district like a district.

Ground-up mixed-use

Seven agents at work

Sixty-seven levers, ten scenarios and seven agents: six composers, each working one part of the deal, and the Anchor holding the feasibility baseline.

Office to residential

Conversions with historic credits

Federal and state historic tax credits sized from the qualified spend, with four specialists for yield, cost, credits and capital, and the market, reasoned live with Claude and checked by the Anchor.

Master-developer district

A district, phase by phase

The strategy for each phase, the value the district creates, and how it flows through the equity waterfall to the institutional partner, the co-GP and the sponsor.

Pre-development capital plan

Before construction starts

The land takedown and the phase-zero budget, and how they are funded before a construction loan is in place.

Built like the software your team already relies on
Pythonengine
Vercelhosting
Next.jsapp
Clerkauth
Neondatabase
Cloudflareedge + files
Claudereasoning
Renderagents
Why it is here

Built for our own deals, shown as an example.

The Upside Explorer came out of our own development work, and it is the clearest example of what we do. Try it on a deal. Then, if it fits, we can go deeper on it with your team, shape it to your work, or build and operate a system to your specifications.

The problem

The upside is real. It is just hard to see.

Industry surveys say nine in ten real estate firms are piloting AI, and one in twenty got what they hoped for; the gap is trust in the numbers. On a complex deal the value hides in the gaps: siloed teams, a pro forma only one person can really drive, and dozens of moving variables that never get tested together. The Upside Explorer makes the whole model live and shared. Change any assumption and watch the deal respond, with your partners and a team of focused agents working the same numbers beside you, every move attributed, and every input labeled as imported, derived or assumed.

How it stays honest

The math is fixed. The intelligence sits on top.

Two parts, kept apart on purpose. One underwriting model does every calculation, the same way every run. The agents only ever propose assumptions. So an agent can be wrong about a number without ever being wrong about the math, and nothing it does can quietly nudge the result.

The model

One model, the same every run

A fixed pro forma that runs the same calculations every time, tested number for number. Same inputs in, same numbers out. No agent can reach in and change it. This is the part you hand to a lender, with no guesswork and no surprises.

The agents

They propose. They never calculate.

A team of agents, each fluent in one part of the deal, propose moves and show their reasoning. A dedicated Anchor holds the feasibility baseline and flags anything that drifts from it. The model does the math.

On a ground-up deal, seven agents act today, each a distinct discipline:

Anchor · Feasibility ProgramCostRevenueOperationsCapital StackRisk & Returns

In development: two policies, an insurance gate and an entitlements agent.

Capital SourcingESG / OutcomesInsuranceEntitlements
How it works

From a baseline to a defensible underwriting, in three moves.

Start from the baseline

Begin with an honest base case, tied to the feasibility study: the program, the costs and the financing it assumes.

Drive it, or hand it over

Move the levers yourself and watch the numbers move with you. Or hand the deal to the agents and watch them find value, one reasoned move at a time.

Get something you can defend

A full underwriting: sources and uses, returns, debt, and program. Behind every assumption sits where it came from and who put it there.

App Launcher →

“An agent can be wrong about an assumption without ever being wrong about the arithmetic. That is the whole design.”

The bigger picture

Two ways a project fails. Both come down to assumptions.

A project can fail because it never pencils. It can also fail a slower way: it gets built, then delivers a place people and the surrounding community did not really need. The first is a financial-assumption problem. The second is an outcomes-assumption problem. Now City is building an engine for each.

Upside Explorerlive

The financial assumptions engine, and the working example you are looking at. It finds the value that makes a project pencil and hold up, with every input labeled by where it came from.

Outcomes ExplorerIn development

The outcomes assumptions engine, and the twin we are building next. It clarifies what a place actually needs, then aligns the upside work here with those outcomes. Read the early thinking →

Bring your next deal to life.

Try it on a deal you know, or book a walkthrough. Then tell us what your team needs: we can go deeper with your team, build to your spec, or drive it with you.

Book a call →Try the live demo