Pointing at Risk: Rates rise further
This should be modeled, but with low probability.
Modeled it: a new Rates rise further event at 10% that adds 0.6 point to the 10-year Treasury yield. The median falls about 0.8% with it in.
extic · Valuation Workbench
Your AI builds the model. You own the odds.
extic turns filings into a probabilistic valuation of every company you own or watch: what's priced in, what could happen, and what would change your mind.
Runs in the Claude or ChatGPT plan you already have. Exploring needs no sign-up.
NVIDIA NVDA Example
Valued 1% under the price
The price needs the guided AI build-out and retained pricing power.
AI build-out · shared with AMD, Arm and Broadcomafter 2027
Works withClaudeChatGPTExcel
How it works
From a 10-Q to the odds behind your position, in one conversation.
Ask Claude or ChatGPT to value a company. It pulls the 10-Q, the call and whatever else it needs, and builds a workbook in which every number points back to where it came from.
| Q3 revenueQ2 call, 2026-08-26 | 108.0B ± 2% |
| Gross margin, fiscal 2028Q2 call, 2026-08-26 | 72–73% |
| Cash and securities10-Q, 2026-07-26, less 1,000 of notes repaid | 55,577 |
A tested engine values every combination of the stories you set odds on. You get a distribution, not one answer.
Example
49.6%of 10,000 futures are worth more than NVIDIA's price
p5 1/3.3median −1%p95 2.1x
Point at what you disagree with, see what the price assumes, and write down what would change your mind. After earnings, one conversation brings it up to date.
- Price needsThe guided AI build-out and retained pricing power
- WatchHyperscaler capex guidance for 2028 · Vera Rubin's share of Data Center
- 2026-10-21Agentic CPUs boom
30%35%, as 2030 server CPU forecasts rose
Question any number
Disagree with a number? Point at it and say so.
Click a number, a story or a paragraph on a workbook or an industry, and speak or type. Your AI knows exactly what you pointed at. Tell it to address your annotations, and it changes the workbook, or tells you why not, and replies with what changed.
Example annotations on NVIDIA's workbook
Hyperscale · Inputs
- Target margin
- 62% ± 3%
- Sales to capital, years 1–5
- 2.5
- Return on capital, 1st terminal period (years 11–30)
35% ± 10%40% ± 5% Addressed
You, pointing at the row
Make this 40 plus or minus five.
You, in Claude or ChatGPT
Address my annotations
Your AI
Changed Hyperscale's return on capital in the moat period from 35% ± 10% to 40% ± 5%. On its own it is worth little: about +0.1% on the median.
Pointing at Book equity · 229,000
Isn't equity 289?
No. 228,984 is total shareholders' equity at 2026-07-26, in the Q2 fiscal 2027 10-Q. Book equity does not move the value.
Who it's for
For investors who do the work themselves.
Independent investors
Twelve holdings. One weekend after earnings. No analyst.
Every company you own gets a model that stays current, with odds instead of one number and a record of why you bought it.
Investor planNVDA1% under
AVGO9% under
AMD49% under
ARM1/4.4 of price
1/10price10x
You build models all day. Your own portfolio gets a spreadsheet from 2019.
A desk-grade model of every holding, with its cost of capital, business lines and claims, that exports to Excel with every cell a formula.
16Free cash flow, year 2=D15*(1-D9)-D1217Discount factor=1/(1+D7)^D318Present value=D16*D1722Value per share212.32
Three partners. One view of the AI build-out.
Set the odds on a shared story once, and every holding it touches moves with them. Roles, and a written rule on who at extic can see your work.
AI build-out · AI computeMRJKAP
MovesNVDAAVGOAMDARM
Why extic
A valuation you can question, because it's yours.
Someone else's fair value can't be questioned, and a spreadsheet gets rebuilt only when there's time. extic's valuations are sourced to filings, tested, and kept current.
Every story has a probability you set.
The page shows what each outcome does to the value, and which stories barely matter.
Custom silicon, NVIDIA's workbook
The return the price implies over your horizon.
Bought at 214.13, NVIDIA's median future returns 12.6% a year over 5 years, just under its 12.7% cost of capital.
cost of capital 12.7%
p5 -17%p95 32%
A tested engine does the arithmetic, not the AI.
Every input cites its filing. Any scenario exports to an Excel workbook whose formulas reproduce the value.
Cash and securities55,577
10-Q, 2026-07-26, less 1,000 of notes repaid in September
Price214.13
Close of 2026-10-16
One conversation brings a workbook up to date.
History shows what changed, when, and why, in your own words.
2026-10-21Agentic CPUs boom 30% 35%
2030 server CPU forecasts rose to $200–300B
2026-10-17Device slump 30% event dropped
"If the build-out falls short" moves all your chips together.
Industries link one factor across workbooks, at one set of odds.
| Median | boom | guided | short | bust | price |
|---|---|---|---|---|---|
| NVIDIA | 332 | 232 | 140 | 73 | 214 |
| Broadcom | 573 | 388 | 216 | 131 | 329 |
| AMD | 349 | 274 | 200 | 127 | 529 |
Runs in the AI plan you already have.
A valuation takes few tokens; ChatGPT Plus is enough. Go further only where a company deserves it.
Example conversation
Update NVIDIA from the Q3 10-Q
Revenue beat the guided 108.0B; I've raised year 1 and cited the filing. Median 212 → 219.
Inside a workbook
Everything a professional model holds, in plain text your AI can read.
Stories and odds
Factors with outcomes, events and spreads, each with your probability.
Implied return
What you earn a year at today's price, over one to ten years.
Cost of capital
Built up from rates, betas and country risk, with every piece shown.
Business lines
Each line its own growth, margins, reinvestment and risk.
Moat duration
How long excess returns last, and on how much new capital.
Claims
Lawsuits, settlements and contingent payments on a timeline.
Buyouts
The chance someone pays a premium before the future arrives.
Banks and insurers
Free cash flow to equity where enterprise value doesn't fit.
Catalysts
Dated events that would settle a story, and what you'll watch.
Industries
Shared factors that keep a sector's workbooks consistent.
HistoryComing
Every version, who changed it, why, and the value at each.
Excel and phone
Formula-driven export, and the full page on a phone.
The library
Open a real valuation. No sign-up.
Public workbooks show their author and date. Explore any of them, change the odds, and see what moves.
Example Industry · four workbooks under one view · 2026-10-16
Open the industry →AI compute
One build-out: NVIDIA and Broadcom priced near value, AMD and Arm far above.
1/101/41/2price2x4x10x
NVIDIANVDA · public workbook
Valued 1% under the price
BroadcomAVGO · public workbook
Valued 9% under the price
AMDAMD · public workbook
Valued 49% under the price
Arm HoldingsARM · public workbook
Valued 1/4.4 of the price
Bars run from p5 to p95 of value against each price, on a log scale; the mark is the median.Shared factors: AI build-out, Custom silicon, Agentic CPUs, Arm vs x86, Taiwan Strait
Pricing
Explore for free. Pay when it's your portfolio.
Free
$0
- Explore every public workbook and industry
- Publish your own valuations
- A few private workbooks to try
Investor Coming
$49 a month, or $490 a year
- Unlimited private workbooks
- Industries across your holdings
- Excel export of any scenario
- History of every change
Team
$120 a seat a month, yearly
- Shared workbooks and industries
- Viewer and editor roles
- Security page and vendor questionnaire
- Onboarding for your team
Paid plans open later. Everyone who joins now starts free.
Privacy and ownership
Your research stays yours.
extic holds research, not client data, and it never tells you what to buy.
Private by default
A workbook is public only when you publish it.
Who at extic can see it
A written rule says who can read your work, and when.
Export at any time
Plain-text workbooks and formula-driven Excel, whenever you want them.
No picks
extic publishes no picks and no fair value of its own. Every valuation is its author's.
For teams
One view of an industry, valued by the whole team.
For emerging managers, family offices and independent advisers, where a few people decide. Add the extic connector to your firm's Claude or ChatGPT organization and keep your AI policy intact.
Talk to us:
Shared industriesEveryone's workbooks move with the team's odds on a shared story.
Viewer and editor rolesAnalysts edit; partners read and comment.
Security pageThe operator-access rule, and a filled-in vendor questionnaire for your compliance officer.
RecordsHistory and plain-text export keep the research behind each decision.Coming
Know the odds behind every position.
Start with one company you own. Your AI builds it from the filings; you set the odds.