Pinned Tweet
๐ H1 2026 is in the books - our first real out-of-sample stretch.
The setup: we built Trade Ideas Flow in late 2025, then froze everything - definitions, methods, hyperparameters, prompts. No tuning. No hindsight. Just the system running daily.
The result:
๐ข TIF: +18%
โช S&P 500: +9%
๐ +9% alpha in 6 months
Every pick was published to the web + emailed to subscribers in real time. The audit trail speaks for itself.
Not advice. Just how we run the process. Past performance โ future results.
Link below if you want to see how it keeps performing and want to get our picks (FREE for now)
4
7
1,419
In June, shareholders of a concrete-machine maker voted down 4 resolutions at its AGM, and the directors owned 0.3% of the stock outright. One of today's writers read that revolt as the catalyst.
๐ Day 2 of the Q3 2026 AUTHOR LEADERBOARD, UNDER 1,000 SUBSCRIBERS. Writers ranked by the average return of their Q3 calls, one call per stock. 6 and 5 went out yesterday:
#6 bought Wolters Kluwer in February and built the valuation model in August to check it
#5 runs a trust that only buys companies with a controlling shareholder
Today: 4 and 3.
1
39
There are 2 ways to top a quarter, and this board has both.
#3 got there on 1 call. Thinkific is up 71.7% and the other 4 calls average about -0.2%, so the median is +4.2% against a mean of +14.2%.
Tomorrow's #2 and #1 got there the other way. They made 12 calls between them and all 12 are up.
I rank by the average because that is what you would have made following every call. The median is what a typical call did, and next week I will post the same 6 ranked that way. 2 of them swap places.
1
16
Tomorrow: a closeout retailer holding 119.3 days of inventory, which its writer argues is the asset, and a model portfolio built 1 sector at a time that made 7 calls in 3 weeks of July.
That was 4 and 3. The top 2 tomorrow.
Today's sources, go support these authors:
#4 @Spread_Hunters
spreadhunters.substack.com
#3 The Stock Stories
thestockstories.substack.com
We read 1,000+ authors and send the best pitches out every morning:
tradeideasflow.ai ๐๐ข
17
For a biotech holder I think this title says everything: "The Science Was Supposed to Be the Hard Part"
$MDNA.TO pitched by David, who built the PRISM-11 framework
Medicenna filed its September accounts on November 16. Next day the stock fell about 37% to C$0.23, more than 80% down over 12 months, about US$15M for the whole company with 4 drug programs inside.
David put the money first, straight from the filings:
โ US$2.6M of cash, about US$475,000 of it in a legal trust account
โ Shareholders' equity negative US$1.5M, new this quarter. It was about +US$1.2M at June 30
โ Working capital went from +US$2.0M in March to -US$1.6M in 6 months
โ Going concern doubt, in the company's own filing
โ A US$5.6M financing agreed in May still has not closed. The filing: "uncertainty remains as to whether the transaction will be completed"
Then the science, which the price says nothing about. MDNA11 is an engineered IL-2, the signal that makes killer T cells multiply. 2 big bets on the same idea failed in 2022, BMS put US$1.85B upfront into Nektar and Sanofi paid about US$2.5B for Synthorx. David's case for why this one differs: binding to CD25, the receptor that lets the cells holding the immune system back grab IL-2 first, is engineered out of the protein itself instead of being covered with PEG, and fusing it to albumin keeps it in the body long enough to work.
In melanoma that already beat checkpoint drugs like Keytruda, median response lasted 68 weeks. One patient had a complete response for 98 weeks, new lesions 33 weeks after stopping, restarted MDNA11 alone and is in a second response 38 weeks on.
Bizaxofusp, for glioblastoma: median survival 13.7 months against 6.2 in a weighted outside comparison group for 19 high-dose patients, and 12.5 against 7.2 across all 44.
And Roche paid US$250M cash in 2022 for Good Therapeutics, whose only program was a preclinical version of the design MDNA113 is built on. About 16 times what market gives all of Medicenna today.
David built PRISM-11 to find out whether the science was good enough to own the stock, and "two years later I think the answer to that question is yes". Still a shareholder, and leaves the call on the price to the reader: "the price is a view about financing and the science is a view about whether any of this works."
1
168
Full piece, the filings line by line and then the case for why this IL-2 differs from the 2 that failed: ๐
davidbakke.substack.com/p/thโฆ
I read every pitch like this at tradeideasflow.ai ๐๐ข
59
๐ The Q3 2026 AUTHOR LEADERBOARD, UNDER 1,000 SUBSCRIBERS! Writers with small audiences whose Q3 calls made the most money for anyone following them.
Before I applied any subscriber cap, the top 3 of the whole Q3 board were already writers this small, so this board is only them.
Ranked by the average return of every long call they made in Q3, which is what you would have made following all of them. One call per stock, their first write-up on it, at least 5 calls each. Marked to Nov 3.
6 made it. Below #6 the averages go negative, and I am not ranking anybody for losing money. #6 and #5 today, and #1 made 7 calls with all 7 up.
1
1
163
5๏ธโฃ @abroninvestor, 762 subscribers
Mean return per call: +6.2% over 9 calls, Q3 2026, marked to Nov 3
Median +5.3%, 8 of 9 up
Best call: $EVT.AX (EVT Limited) +23.8%, A$12.53 to A$15.51
Andrew Brown's EVT write-up: the Rydge family controls 42%, the hotels are worth more than the cinema chain, and a ski field lease to 2057 and a Sydney CBD site sit on top. Shares were at the level of 11 years ago. "In our view the shares have a midpoint value of over $19 before tax."
The biggest day since was Aug 24, when EVT put about A$800 million of non-core property up for sale and hired Rothschild to review the group structure.
Brown runs a trust that only buys companies with a controlling shareholder, and the 9 calls read like it: Exor at about 57% below the value of what it owns, HAL, E-L Financial, D'Ieteren, CK Hutchison. Listed in 8 countries, none in the US.
1
1
51
Tomorrow: a London-listed maker of concrete-levelling machines whose shareholders voted down 4 resolutions at the AGM, and a write-up that took C$1.05 of net cash out of each C$1.34 share and found the rest of the business selling for C$0.29.
That was 6 and 5. Tomorrow: 4 and 3.
Today's sources, go support these authors:
#6 @CompoundingLab
compoundinglab.substack.com
#5 @abroninvestor
andrewbrown174150.substack.cโฆ
We read 1,000+ authors and send the best pitches out every morning:
tradeideasflow.ai ๐๐ข
1
55
Best opening line I read today: "I won a tin trophy, so I went looking at a tin company."
$TORCF pitched by 8020Invest
8020Invest won the award for predicting the tin price at the LME Dinner Downunder in Melbourne, US$51,100 on the trophy. Then at the Canaccord Noosa mining conference caught up with Rob Curtis, now running TinOne Resources. "So I went home & read everything I could find."
TinOne explores for tin and tungsten in north-east Tasmania, listed in Canada on the TSXV as TORC rather than on the ASX. About C$5.5M for the whole company at 15c, and it closed at C$0.14 yesterday.
"His experience is the reason I'm looking." Geologist, 30+ years, Rio Tinto, Oxiana, OZ Minerals, EMR Capital. He was on the early Oxiana team that bought the Sepon project when Oxiana was worth about $30m. That company became OZ Minerals, which BHP bought for about $9.6b. CEO of TinOne since 8 September.
"I was a CPA, so I went to the accounts first."
โ At 31 March: C$40,432 of cash, C$846,389 owed, C$556,625 of it to directors and their companies, going concern warning
โ Since Rob arrived: C$750,000 raised in October at 5c, insiders put in C$185,000 of it
โ C$172,137 of debt being swapped for shares at 14c
โ 9.3M shares in June 2025, 36.9M now
THE ROCK
Great Pyramid holds 8.39M tonnes at 0.17% tin, 14,400 tonnes of tin, in the lowest confidence category. 2024 drilling put 5 holes beside holes from 1970 to check the old numbers. Best one returned 29m at 0.71% tin, including 14m at 1.03%. 2 holes within a metre or 2 of the old ones came back with about 70% and 21% more tin, and 1 found nothing.
That resource was worked out at US$24,978 a tonne. Tin is close to double that today. It also assumes 80% of the tin can be recovered, a number from 1980s test work, so the new recovery tests are what 8020Invest cares about most. Drilling was due to start in November, 17 holes across Aberfoyle, Royal George and Great Pyramid.
No position, and it barely trades, which 8020Invest really does not like. "I can see a path to 2 or 3 times if the first results are good & the next raise is done at a higher price."
1
139
Full write-up, from the trophy to the 1980s recovery tests, with what the next 3 announcements need to show: ๐
8020invest.substack.com/p/i-โฆ
I read every pitch like this at tradeideasflow.ai ๐๐ข
49
I like a forced seller with a date on it, and this one has December 11.
$HYPR pitched by @kankanandb
Hyperfine makes Swoop, a portable MRI scanner. Brian, who writes the piece, calls it the only FDA-cleared portable low-field MRI on the market.
Company was added to the Russell indexes in June. On November 13 it was on the preliminary deletion list, so funds that own it only because of the index have to sell by December 11, whatever the business does. Brian puts the fall from $0.72 at earnings to $0.56 on that selling, plus traders getting in front of it since late October.
So what does $0.56 buy. Brian did the arithmetic off the 10-Q.
THE NUMBERS
โ About 106.2M shares, so about $59.6M for the whole company
โ $37.9M of cash at September 30, about $0.36 a share
โ Take the cash out and the business itself is priced at about $21.7M
โ $25M of credit from Horizon Technology Finance, undrawn, available through end of 2027
โ Stock closed $0.562 yesterday
THE QUARTER
Q3 revenue grew 73% on a year, which management called organic with no one-time items. Gross margin was a record 56.7%, fifth quarter in a row above 50%. Cash burn came down from $8.8M in Q1 to $5.6M in Q3, and on management's own guidance the cash lasts into 2028.
Full year guidance of $20 to 22M was reaffirmed. Hitting the low end needs about $6.2M in Q4, roughly flat on the quarter just reported.
Next real catalyst after the index is contrast imaging, with the FDA submission on track for year end and clearance aimed at mid-2027.
Brian owns the shares. "December 11 is when the forced selling clears. I am watching that date."
$HYPR trades at $0.56. Cash on hand alone is worth $0.36 per share. The market is pricing the entire business โ FDA clearances, installed base, contrast catalyst, $6M record quarter โ at $21 million net of cash.
Short sellers are borrowing against a clearance sale that they front-ran.
December 11 the index selling clears. New article: takedown2021.substack.com/p/โฆ
#HYPR #Hyperfine #FinTwit #FinX #SmallCaps #MedTech #ShortSqueeze #Oversold #ValueInvesting #Russell2000 #IndexRebalance #ShortSellers
1
1
149
I read a short report today that starts with a logo.
$NCPL pitched by @fugaziresearch
Netcapital runs a crowdfunding platform. In August the SEC sued the company and 5 people around it, alleging about $13.9M of revenue came from sham consulting deals. One example from the complaint: a startup agreed to pay $1.3M for 6 services including logo design and a website. Its website and logo stayed the same from at least 2010 to May 2025.
Normal clients paid about $50,000 for that kind of work. The SEC says at least 11 of these customers were controlled by John Fanning, whose spouse was the CFO, and they paid in their own shares rather than cash.
Fugazi Research did one sum of its own. $13.9M is about 77% of the $17.95M Netcapital reported over those 10 quarters.
2 days after the complaint the auditor resigned. Company told investors not to rely on its audited accounts for fiscal 2022 to 2025, a new auditor is redoing all of it, and Nasdaq gave it until February 9, 2027 to file the missing reports.
FIX THE BOOKS, TRIGGER THE DEBT
Then Fugazi went through the October note from Hinterland Partners. Lender can turn the debt into shares at 15% below the lowest price of the previous 7 trading days. And the note lists a qualifying restatement of the accounts as an event of default, the lender's right to call the loan. So the re-audit it needs to file its reports and stay listed can hand the lender a default.
THE NUMBERS
โ Last filed 9 months, to January 31: revenue $335,481 against $7.90M of operating expenses, $23.56 spent for every $1 earned
โ Shares went from 2,192,226 to 6,847,899 in those same 9 months
โ At least 7 discounted note financings since May, on top of a $15M equity line
โ Stock closed $2.37 yesterday. Fugazi counts a move of more than 1,000% from $0.21 since late August
"Fugazi Research considers NCPL uninvestable at any price above zero."
$NCPL: The SEC charged it with a fraudulent scheme. It says $13.9M of revenue was sham, a 345% overstatement, with some contracts forged.
The auditor has quit. Four years of audited financials are now โdo not rely.โ
And the stock is up 1,000% from $0.21.
Fugazi Research considers NCPL uninvestable at any price above ZERO.
Click for our full report โฌ๏ธโฌ๏ธ
fugaziresearch.com/p/ncpl-phโฆ
1
166
๐ THE LARGEST SHAREHOLDER JUST CHANGED
$JCTC pitched by Terry Schumacher
Adam Bradley of AJB Capital now owns 564,321 shares of Jewett-Cameron, 16.03% of the company as of 11/16. That makes him the largest holder. He started buying in March, and his SEC filings say only that he thought the shares were "undervalued".
Why this matters here.
"Jewett-Cameron is failing as an operating entity."
8 straight fiscal quarters of losses through 5/31/26, -$10.3M net income and -$9.1M EBITDA added up. Terry estimates the 2 quarters since lost money too, which would make 10.
At most companies board fires the CEO after that. Here company spent an estimated $1.4M on consultants over 3 quarters, with no explanation of how it helped, and losses kept coming. Management and directors, except director Scott Kotarba, own very little stock.
THE NUMBERS
โ Price $2.88
โ Terry's liquidation value $6.75 a share, meaning sell the real estate and other assets, pay every liability, and that is what is left per share
โ +134% from here
โ Bradley's AJB Investment Fund II: 19.8% a year net since it started in 2007
THE CATALYST
8/31 quarter results due close to 11/30, the 11/30 quarter close to 1/19/27. Company usually files right on the deadline.
Terry owns 50,000 shares since last December and expects Bradley to use the new position to push for governance change.
"I am looking forward to the opportunity to vote my shares in favor of a completely new board of directors."
1
142
Terry's full piece, with Bradley's track record and the case for a new board at Jewett-Cameron: ๐
terryschumacher.substack.comโฆ
I read every pitch like this at tradeideasflow.ai ๐๐ข
51
๐ MY FAVORITE STOCK PITCH SO FAR THIS WEEK
$032560.KS pitched by @RogersDeepValue
Korean company worth about KRW 92.6bn on the market. Its net current assets, everything it owns short term minus every debt and minus what outside holders own of its subsidiaries, were KRW 145.8bn at end of September. Most of that is stainless stock in the warehouse, but still.
Hwang Kum Steel & Technology. Cuts, polishes and sells stainless steel south of Seoul for POSCO and others, owns two-thirds of a listed pipe maker and most of a bridge builder. Made a profit every year from its 1997 listing to 2023.
So why so cheap. Chairman Kim Jong-hyun and family own 56.42%. Payout is about 14% of profit and there is no return policy. When 2 small holders proposed a dividend and a value-up plan in Feb 2025, board kept it off the agenda because it came after the legal deadline. And some cash went to family ventures, like KRW 4.0bn for 83.33% of a farming company chairman set up in 2019.
"The market prices Hwang Kum as if the family will sit on the balance sheet forever."
The part I liked is the work on the cash. Group reports KRW 86.5bn of net cash, but most of it sits in subsidiaries that have their own outside holders, and at the bridge builder some of it is customer prepayments. Rogers Deep Value went through the Korean filings and took all of that out, which left about half for Hwang Kum's own shareholders. Then the Esang stake counted at half its book because the market values it at less than half, plant and land at half, and 30% off everything for family control.
"Our answer: most of that is true, and we have priced it."
THE NUMBERS
โ Price KRW 5,450
โ Value after all those haircuts: KRW 12,014, +120%
โ If family ever shares, no control discount: KRW 13,970, +156%
โ 9 month operating profit KRW 30.5bn against KRW 13.3bn a year before, filed Nov 12
โ Shareholders' own net cash up from KRW 45.2bn in June to KRW 54.9bn in September
THE CATALYST
In March Korean regulators said they will publish lists of companies stuck in the bottom fifth of their industry on price to book, sparing those that file a value-up plan. First lists reported for this month. Hwang Kum has filed no plan.
Family's other listed companies already moved. USTeel published a value-up plan and paid out 57.7% of its 2025 profit, Esang is cancelling 3.69% of its shares. And the chairman bought 390,376 more shares on the market between April 2025 and January, with his own money.
"I'm buying this. Not advice."
1/ new write-up: Hwang Kum Steel, a Korean stainless steel company I own. current assets minus every liability and the outside holders' share of its subsidiaries: โฉ145.8bn. the market pays ~โฉ93bn. plants, land and a listed stake for free ๐คฏ
I'm buying this. Not advice.
137
Most interesting thing I read this week is about phone software for small businesses.
$CXDO pitched by @sammccolgan
Crexendo. Around 250 local telecom and IT providers rent its platform and sell phone service under their own brand, "much as an online shop runs on Shopify without its customers knowing". Over 8 million people make work calls on it, most never heard the name.
Stock has halved since June and closed Friday at $4.92. Every bear Sam talks to says the same thing, software growth is slowing. And licence sales did fall 25% in the first 9 months. Licences are the lumpy part though, capacity bought upfront in blocks, and new customers now start with smaller orders and top up as they move users across. The recurring line underneath grows 16 to 20% a year. Hold licences flat and software growth is back to about 13% in 2027 by Sam's maths.
For proof Sam went to BroadSoft, the incumbent in this exact market until Cisco bought it in 2018, which reported the same 2 lines for a decade. Its licence growth swung from -5% to +70%, back to roughly zero, up to 24%, down to -3%. Recurring line grew every quarter.
"The market is busy counting Crexendo's licence sales. I think it is missing the gorilla."
THE GORILLA
Management dropped pieces of it on calls and at conferences: what the AI receptionist earns per account, how many licensees resell it, a price range for call analysis, how fast the app store grows. "I have yet to see anyone put the pieces together." Sam did, and counts 5 income streams.
The number that matters. Crexendo makes about $80 a year from each business on a licensee's platform, and there are about 390,000 of them. CAIRO, the AI receptionist, earns it about $900 a year from each business that takes it.
THE NUMBERS
โ EBITDA (roughly cash profit before interest and tax) about $15M this year
โ Sam's base case: about $54M in 2028, about $40M if AI take-up is half the assumption
โ Analysts: $21M
โ Enterprise value about $150M at the $5.07 it was written at, so the price pays for almost none of it
THE DEALS
Spare cash goes into buying its own licensees at about 1.3x revenue. Same platform, so no migration risk. ESI cost about 10x EBITDA, about 8.4x after savings, and about 5.5x once Crexendo sells the AI add-ons to the customers it bought.
A year from now Sam values it at about $6.40 with no AI and no deals, $8.10 with the deals, $18 to $24 with the AI streams. First real test is year-end results, when management gives CAIRO numbers for the first time.
Sam and the fund are long.
"If I'm wrong, I own a growing, cash-generative business near its historical lows. If I'm even half right on AI, the shares are worth several times today's price."
Sam's full write-up, with all 5 AI streams modelled line by line and every assumption on the page to poke holes in: ๐
cadoccapital.substack.com/p/โฆ
I read every pitch like this at tradeideasflow.ai ๐๐ข
93











