Given a project, write an investment analysis. The entire report answers only one question: is this thing creating a new order, or is it moving the old order around.
Foundation
A single definition holds up the entire report: wealth is not money; it is order illuminated by desire, and money is merely the unit of measurement for order. Investing is taking the order in your hands and exchanging it for a better order generator.
Starting from this definition, the report's questions differ from market convention:
| Conventional question | This report asks |
|---|---|
| How much is this company worth | Can this machine run |
| How big is the market | What outdated labels is the market using to see it |
| How much can it rise | What am I giving up for what, and after the exchange, who is smarter |
Input
Company name, BP, written introduction, conversation records—any material describing the project will do. For a well-known company, just the name is enough—use the Research skill or a subagent to pull the latest financial reports and industry data; do not write from old impressions.
Report structure
The five sections are a skeleton, not a fill-in-the-blank exercise. Whichever section has the most material for this project, write more there; for one with no material, cover it in a sentence or two, or skip it entirely. The report serves the judgment; completeness for its own sake is meaningless.
I. What is this
A table, plus a track definition you write yourself.
| Dimension | Content |
|---|---|
| Project name | |
| Track definition | In our own words, do not copy market labels |
| Stage | |
| Financing situation | Amount / valuation / terms (fill in if available, note if not) |
| Data snapshot | Key operating data |
The track definition must state what this company is truly doing—the layer that market labels cannot express. Calling it a "search engine company" says nothing; saying it is "a monopoly operator of humanity's cognitive infrastructure" puts what it makes money from and what it fears all in that one sentence.
II. Judgment as an order-creation machine
The weight of the entire report rests on this section. Do not score item by item; answering one question is enough:Can this machine run? Look from three angles.
Is the flywheel turning? First find the loop in the system that gets better the more it is used: more users means more data, more data means a better product, a better product means more users. Once found, see what step it has reached—stopped, just starting, already turning; if turning, see how long it has been turning and whether it is accelerating or flattening; if stopped, write out the link that is stuck.
After a shock, does it become stronger or weaker? When competition cuts in, technology changes generations, or the market collapses, will this machine shatter, hold up, or eat the shock as its own fuel. Go through its history: has it taken hits, and after taking hits did it weaken or grow stronger.
Are resources pushed to it, or do they come on their own? Expansion by negotiating one by one and buying piece by piece—that is pushing. When others come flooding in on their own and staying away means losing out—that is gravity. Look for signs of "gathering without pushing."
Overall judgment, choose one of three tiers:
- Order-creation machine—the flywheel is turning, it gets stronger after shocks, resources come on their own
- Has potential—the flywheel's structure is there, but it hasn't been verified to be able to spin
- Order-transfer—rearranging what already exists, without generating new order
III. Creation formula
Every order-creation machine has a core algorithm; write it out in one sentence. Reference:
- Amazon = profit → reinvestment → lower costs → lower prices → more users → more profit
- Tesla = hardware collects data → data trains algorithms → algorithms redefine hardware
- Google = each time the way humans find answers migrates, become the default infrastructure for the new way
After writing it, ask two further questions: how many times this formula has been verified, and to what degree; whether anyone else is running a similar formula, and where the difference lies.
IV. What the market sees vs. what we see
Investment timing is decided in this section.
Where is it on the S curve? Accumulation phase, inflection point, acceleration phase, plateau phase—which segment does it fall in. For those before the inflection point, write clearly what conditions will trigger the inflection point.
What old eyes is the market using to see it? What label has the market stuck on it, what does this label obscure, what does our framework see more of. How large this cognitive gap is—excess returns come from here. For cognitive discount, look at three signals: it takes great effort to explain it before others understand; pricing is anomalously off for a long time, and the parts added together do not equal the whole; all ready-made analogies fail to match—it is like X yet not like X.
What has it controlled that others cannot take away? What it holds is data, distribution, standards, or network effects. Whether this control is static (brand, patents) or grows stronger over time. Look one step further: will this scarcity move elsewhere in the future, and can the project keep up.
Which free ride is it taking? Three kinds of cost are collapsing: understanding cost, collaboration cost, action cost. Which one is this project riding, and how much of the energy released by the collapse has it caught.
V. Swap or not
- Exchange recommendation: recommend invest / recommend observe / recommend abandon
- If investing: recommended amount range, key terms
- Core assumptions: which assumptions this decision is betting on. Pair each assumption with an exit signal—what data appearing means the assumption is wrong and it is time to leave.
- Unresolved questions: 3-5 questions that matter to the decision but have no answer yet, ranked by importance.
Final sentence
Answer in one sentence: what is the essence of this project—creating new order, or transferring old order.
Output
- Format: org-mode
- Directory:
~/Documents/notes/ - Named by denote:
YYYYMMDDTHHMMSS==z--投资分析-PROJECT_NAME.org, such as20260326153000==z--投资分析-example-ai.org - Write it using the Write tool, and after writing report the full path to the user
Generation rules
- Use only real information. Mark directly anything that cannot be found; do not force it, do not make it up.
- Dare to make judgments. Not a single sentence like "it could be good or bad" is allowed.
- Each judgment carries evidence: data, citations, specific facts.
- These words are prohibited: huge market, excellent team, broad prospects, blue ocean market.
- The length should be based on what is needed to explain it thoroughly. If two thousand words can make it clear, then two thousand words; if seven thousand words are needed, then seven thousand words.
- Write in Chinese.