Research that cannot rewrite its record.
We publish the exact price at which the math starts working for each company we cover — then grade the call in public, forever. Losers included.
Deep research, cited
Twelve companies across three emerging industries — AI infrastructure, nuclear fuel, space. Every number traced to an SEC filing you can open.
A clear price for each
Not a vague rating — a specific acts-below price where the math finally works. Above it we wait, publicly, no matter how loved the stock is.
A record we can't edit
Every miss stays on the board next to every hit, permanently — and you can check that claim yourself rather than take our word for it.
Every stock-picking shop shows you its winners.
The losers vanish quietly. The windows get cherry-picked.
“Since inception” gets redefined until it flatters.
So we built a record that makes hiding impossible — for us.
Published once. Provable forever.
Three rules make the record impossible to fake — each one enforced in code, not policy.
The thesis is sealed
Scenario ranges, every assumption, and pre-registered kill conditions are hashed and signed into the chain the moment a call goes live. Revisions can only be appended — never swapped in.
chain_hash = sha256(prev ‖ payload ‖ type ‖ ts ‖ seq)No lookahead, by construction
Entry is the first market close after the publication timestamp — never before, never same-day. Every call is benchmarked against QQQ over the identical window.
entry = first_close(date > t_publish)The board never forgets
Quarterly reviews are signed events. A triggered kill condition forces a close with a post-mortem. Every call ever published stays on the board, for good.
α = r_call − r_QQQ, identical windowBuilt from what survives the evidence.
We went through the research — the factor literature, the base-rate studies, the records of the best long-term investors — and adopted only what held up. Every rule cites its source on the methodology page; this is the shape of it.
Plain reading — thousands of companies go in; twelve earned research; none have cleared the price gates yet. The zero isn't failure — it's the discipline holding.
Compounders
About 4% of stocks have created essentially all market wealth over cash — and history's greatest of them returned roughly 13.5% a year, held for decades. The edge isn't finding rockets; it's owning elite, persistently profitable businesses at defensible prices and then not flinching. Profitability persists. Growth forecasts don't. We underwrite the first, never the second.
Moonshots
Real 10-baggers didn't start as lottery tickets — the median began as a ~$350M company at 0.6× sales, already profitable. Lottery traits are a measurable tax, so each one is a veto here: no penny stocks, no spike-chasing, no serial diluters. And moonshots only work as a basket — small positions, many names, ten-year holds, drawdowns expected.
Concentration with conviction has evidence behind it; concentration without rules is how the math kills you. Our published construction method: no name above 10% at cost, large positions capped in aggregate, no more than a quarter of the book in one correlated cluster, moonshots sized like venture bets, everything anchored at a fraction of Kelly because estimation error is real. Sell rules are sealed before the buy case can be.
This is what refusing $339 looks like.
The full Celestica research is sealed at seq 1 of the chain — a business we admire, at a price our math can't defend. These are its actual published numbers: the model, the range, the grid, the simulation — and the two gates it failed. Read the sealed original →
Plain reading — our entire valuation method is this one line. Nothing you can't check — you could redo it on paper in a minute.
Plain reading — every outcome we can defend is worth less than the $339 being asked. So we published the research — and wait at our line.
Plain reading — each cell is what the stock is worth under different growth and exit assumptions. Only the most heroic corner justifies today's price.
Plain reading — we simulated 10,000 possible futures using our own published odds. Roughly three in four end below today's price.
Plain reading — two hard tests decide every call. Celestica fails both at today's price — so it's a Watch, not an open thesis.
Every call lives on a clock.
Publication seals it. The next market close prices it — never the same day, never before. Every quarterly filing re-tests it against its own kill conditions, year one grades the published range, and year five delivers the verdict. All of it signed. None of it optional.
Plain reading — every call is re-tested against its own kill conditions each quarter and graded for five years. Nothing gets quietly forgotten.
The scoreboard, losers included.
Forward-only from genesis (2026-07-23). We researched twelve of 2026's most-loved growth stories at launch; at today's prices, none cleared our valuation gates — so every one is a sealed watch call with the price at which our math starts working. If a stock we refused beats the index anyway, the record will say our discipline cost money.
Sorted by distance to our line. When a price crosses it, the gates re-run — and a Watch can become an open thesis, signed like everything else.
MOD
Acts below $166.19
Now $201.06 · +21% above
LEU
Acts below $101.65
Now $176.93 · +74% above
BWXT
Acts below $93.44
Now $168.70 · +81% above
FLY
Acts below $9.95
Now $20.33 · +104% above
LUNR
Acts below $5.82
Now $12.34 · +112% above
CRDO
Acts below $90.24
Now $206.99 · +129% above
CLS
Acts below $128.42
Now $331.44 · +158% above
PL
Acts below $7.82
Now $20.48 · +162% above
CCJ
Acts below $32.21
Now $86.38 · +168% above
RKLB
Acts below $13.01
Now $64.95 · +399% above
UEC
Acts below $1.90
Now $9.60 · +405% above
ALAB
Acts below $25.85
Now $311.23 · +1104% above| Company | Call | Published | At price | Acts below | Last | Return | vs QQQ |
|---|---|---|---|---|---|---|---|
| ALABAstera Labs, Inc. |
Watch | 2026-07-23 | $326.97 | < $25.85 | $311.23 | +6.7% | +6.2% |
| BWXTBWX Technologies, Inc. |
Watch | 2026-07-23 | $176.67 | < $93.44 | $168.70 | −3.3% | −3.9% |
| CCJCameco Corporation |
Watch | 2026-07-23 | $90.37 | < $32.21 | $86.38 | −1.7% | −2.2% |
| CLSCelestica Inc. |
Watch | 2026-07-23 | $339.10 | < $128.42 | $331.44 | +8.6% | +8.0% |
| CRDOCredo Technology Group Holding Ltd |
Watch | 2026-07-23 | $236.50 | < $90.24 | $206.99 | −2.9% | −3.4% |
| FLYFirefly Aerospace, Inc. |
Watch | 2026-07-23 | $20.70 | < $9.95 | $20.33 | +3.2% | +2.7% |
| LEUCentrus Energy Corp |
Watch | 2026-07-23 | $170.78 | < $101.65 | $176.93 | +8.0% | +7.4% |
| LUNRIntuitive Machines, Inc. |
Watch | 2026-07-23 | $13.71 | < $5.82 | $12.34 | −4.5% | −5.0% |
| MODModine Manufacturing Company |
Watch | 2026-07-23 | $249.62 | < $166.19 | $201.06 | −16.9% | −17.4% |
| PLPlanet Labs PBC |
Watch | 2026-07-23 | $22.36 | < $7.82 | $20.48 | +0.1% | −0.5% |
| RKLBRocket Lab Corporation |
Watch | 2026-07-23 | $67.28 | < $13.01 | $64.95 | +1.6% | +1.1% |
| UECUranium Energy Corp |
Watch | 2026-07-23 | $9.73 | < $1.90 | $9.60 | +0.9% | +0.4% |
How to read it: Watch = right business, wrong price. Acts below = the price where our math clears a 15%/yr bar. If a stock we refused beats QQQ anyway, the board shows what our caution cost.
What the tape says. For context.
12-1 momentum is the one technical factor that survived a century of academic scrutiny — and, more to the point, our own testing, where nearly everything else died. So we track it for every name on the board you just read, and we let it inform exactly nothing: valuation gates decide, momentum watches.
Plain reading — which of our names are in strong uptrends right now. Useful context — but price versus value decides, never the trend.
Plain reading — the dashed lines are where chart traders expect support. The solid one is ours. We only act on the line with arithmetic behind it.
The record fits in a pocket. So does the proof.
The whole registry is one JSON file and an open verification spec. Read it on a phone, replay it on a laptop, mirror it anywhere — it verifies the same everywhere, because it's math, not trust.
The industries being built now.
A theme publishes only when the inflection is already in the filings. One exhibit per theme, from the sealed research — the rest lives on each research page:
MODData-center cooling +73% to $1.1B, with a $4B capacity agreement for 2027–29FY26 10-KLEU$3.9B enrichment backlog contracted out to 2040Q1 FY26 10-QRKLBRevenue +63.5% y/y with backlog of $2.2B, up 108%Q1 FY26 10-QAI Physical Infrastructure
Power, cooling, and interconnect — the physical layer being rebuilt under the AI datacenter boom, bought through the suppliers rather than the obvious mega-caps.
4 CALLS → Theme 02Nuclear Fuel Cycle
The Western nuclear fuel chain restarting after 30 years of atrophy — enrichment, components, and uranium, where the backlogs already run to 2040.
4 CALLS → Theme 03Space Economy
Launch-cost-curve beneficiaries with actual revenue — the companies whose order books inflect every time cost-per-kilogram falls.
4 CALLS →