Question for the group:
With AI advancing so quickly, how are you using it in your stock investing/research?
Curious to hear what tools and workflows people actually use.
I know the community actually puts in the fundamental work - reading 10-Ks, building DCF models, and ruthlessly analyzing moats. But I’m curious about your day-to-day realities.
I recently sat down to map out my own investing "process" and realized my honest, real-world version is embarrassingly short compared to the ideal.
My reality: Look at the P/E and FCF yield, skim the most recent earnings transcript, read a couple of write-ups or Reddit comments, and hit buy. Maybe 10 to 20 minutes total.
The "ideal" value version in my head - combing through 5 years of 10-Ks, building a rigorous DCF from scratch, calculating ROIC variations, tracking insider buying, and patiently waiting for a margin of safety -rarely happens as thoroughly as I plan it to.
Since I know you guys dig much deeper, I want to know what your actual workflow looks like... how you manage your time and research:
Everyone spent Friday celebrating Amazon's blowout quarter. Stock up 15%, AWS growing 37%, net income tripled to $62.6 billion.
Almost nobody read the footnote.
$53.4 billion of that "income" is a paper gain on Amazon's stake in Anthropic. It's right in the press release: "non-operating other income... primarily from investments in Anthropic." Amazon's actual business - every warehouse, every Prime sub, all of AWS - earned $27.5 billion.
The markup on one AI startup made Amazon twice as much as Amazon.
And here's where it gets circular... Anthropic is one of AWS's biggest customers -->
Amazon invests in Anthropic -->
Anthropic spends a chunk of it back on AWS compute. AWS accelerates to its fastest growth in 18 quarters. And on top of that, Anthropic's valuation jumps, so Amazon marks up its stake and books $53 billion of income it can't spend... It's not cash. Nobody sold anything. The number just went up!
To be fair, none of this is hidden or improper - it's mark-to-market accounting doing what it does. And it cuts both ways: if Anthropic's valuation ever drops, the same line shows up as a giant paper loss in some future quarter.
Everyone spent Friday celebrating Amazon's blowout quarter. Stock up 15%, AWS growing 37%, net income tripled to $62.6 billion.
Almost nobody read the footnote.
$53.4 billion of that "income" is a paper gain on Amazon's stake in Anthropic. It's right in the press release: "non-operating other income... primarily from investments in Anthropic." Amazon's actual business - every warehouse, every Prime sub, all of AWS - earned $27.5 billion.
The markup on one AI startup made Amazon twice as much as Amazon.
And here's where it gets circular... Anthropic is one of AWS's biggest customers -->
Amazon invests in Anthropic -->
Anthropic spends a chunk of it back on AWS compute. AWS accelerates to its fastest growth in 18 quarters. And on top of that, Anthropic's valuation jumps, so Amazon marks up its stake and books $53 billion of income it can't spend... It's not cash. Nobody sold anything. The number just went up!
To be fair, none of this is hidden or improper - it's mark-to-market accounting doing what it does. And it cuts both ways: if Anthropic's valuation ever drops, the same line shows up as a giant paper loss in some future quarter.
Everyone spent Friday celebrating Amazon's blowout quarter. Stock up 15%, AWS growing 37%, net income tripled to $62.6 billion.
Almost nobody read the footnote.
$53.4 billion of that "income" is a paper gain on Amazon's stake in Anthropic. It's right in the press release: "non-operating other income... primarily from investments in Anthropic." Amazon's actual business - every warehouse, every Prime sub, all of AWS - earned $27.5 billion.
The markup on one AI startup made Amazon twice as much as Amazon.
And here's where it gets circular... Anthropic is one of AWS's biggest customers -->
Amazon invests in Anthropic -->
Anthropic spends a chunk of it back on AWS compute. AWS accelerates to its fastest growth in 18 quarters. And on top of that, Anthropic's valuation jumps, so Amazon marks up its stake and books $53 billion of income it can't spend... It's not cash. Nobody sold anything. The number just went up!
To be fair, none of this is hidden or improper - it's mark-to-market accounting doing what it does. And it cuts both ways: if Anthropic's valuation ever drops, the same line shows up as a giant paper loss in some future quarter.
Reddit just had maybe the best quarter of any social platform this year... revenue up 61%, profit almost tripled, users up 18%.
Stock dropped 20% anyway (it was down 23% at the lows, per CNBC).
Why? Four words buried in the investor letter: "search referrals were choppy."
That's it. That's the crash.
Basically most new people find Reddit by googling stuff.
And Google's AI now just answers the question right there on the results page, so nobody clicks through... so Wall Street ignored the great numbers and focused on one fear: if Google stops sending people, where does Reddit's growth come from?
And here's the part that actually breaks my brain: Reddit sells its content to Google to train that same AI... Google pays Reddit for the data, then uses it to answer questions so people don't visit Reddit.
The CEO got on the call and said people don't want AI summaries they want Reddit. The market listened to that argument and took the stock down 20%.
source:
https://www.cnbc.com/2026/07/31/stocks-making-the-biggest-moves-midday-aapl-amzn-rddt-gddy-iesc.html
Killed from both sides: long the AI buildout (SK Hynix, Nebius, SanDisk, Micron, CoreWeave - all down 35%+), short software names like Adobe that ripped higher.
The part that should scare you: forced sellers don’t care about your cost basis. They were forced to sell off their entire public portfolio in one shot. Wednesday’s 1,100-point Dow drop was probably partly this book getting liquidated into no bid. Notice how chips bounced today the moment it cleared?
His letter says the fund was “not immune” but calls this the best opportunity since early 2025 - and actually invites investors to add capital Aug. 1. a bold move!
Killed from both sides: long the AI buildout (SK Hynix, Nebius, SanDisk, Micron, CoreWeave - all down 35%+), short software names like Adobe that ripped higher.
The part that should scare you: forced sellers don’t care about your cost basis. They were forced to sell off their entire public portfolio in one shot. Wednesday’s 1,100-point Dow drop was probably partly this book getting liquidated into no bid. Notice how chips bounced today the moment it cleared?
His letter says the fund was “not immune” but calls this the best opportunity since early 2025 - and actually invites investors to add capital Aug. 1. a bold move.
Killed from both sides: long the AI buildout (SK Hynix, Nebius, SanDisk, Micron, CoreWeave - all down 35%+), short software names like Adobe that ripped higher.
The part that should scare you: forced sellers don’t care about your cost basis. They were forced to sell off their entire public portfolio in one shot. Wednesday’s 1,100-point Dow drop was probably partly this book getting liquidated into no bid. Notice how chips bounced today the moment it cleared?
His letter says the fund was “not immune” but calls this the best opportunity since early 2025 - and actually invites investors to add capital Aug. 1. a bold move.
The trigger is almost absurd. SK Hynix posted the most profitable quarter in Korean corporate history - 60.5 trillion won in operating profit, up over 550% (Bloomberg) - and it still missed estimates. The stock fell about 10%, the index dropped 8%, and Korea got its first back-to-back circuit breakers ever. July is now the worst month in KOSPI history.
But the earnings aren't the real story. The leverage is. Retail investors lost about $1.5 billion in those nine-week-old ETFs, and over 320,000 accounts were force-liquidated (IBTimes). The 2x SK Hynix product is down 80% from June while the stock itself is down far less. That's daily-reset decay: the underlying can recover, the 2x holder doesn't.
Nobody cut AI demand forecasts today. A company growing revenue 257% got sold anyway.
The US lists the exact same product structure.
And Microsoft and Meta report tonight into this exact mood....
Is Seoul just a leverage flush, or the first crack in the AI trade?
China’s DRAM giant, Changxin Technology Group (CXMT), made headlines with a massive 466% surge on its Shanghai STAR Market debut, raising $8.6 billion and briefly becoming China’s most valuable listed stock at a $487 billion market cap. For those watching the semiconductor space this eye-popping first-day gain was largely a classic low-float squeeze with only about 7% of shares available to trade driven by heavy state support for chip self-sufficiency and reports that Apple is testing CXMT’s DRAM for local Chinese devices. The company backed up the hype by swinging to a 35.4 billion yuan Q1 profit, benefiting directly from the ongoing global AI infrastructure boom.
while leaders like Micron ($MU) and SK Hynix are shifting cleanroom capacity to high-margin high-bandwidth memory (HBM) for AI, CXMT is aggressively expanding standard DDR5 and mobile memory, targeting 350,000 wafer starts per month by year-end.
However, trade restrictions prevent CXMT from acquiring advanced Extreme Ultraviolet (EUV) tools, leaving its production costs roughly 30% higher than Western peers. Ultimately, chasing low-float, first-day pops near the top of a memory cycle carries severe downside risk. Rather than buying into the hype, retail investors should monitor how CXMT's massive supply expansion impacts global DRAM pricing over the next two years.
China’s DRAM giant, Changxin Technology Group (CXMT), made headlines with a massive 466% surge on its Shanghai STAR Market debut, raising $8.6 billion and briefly becoming China’s most valuable listed stock at a $487 billion market cap. For those watching the semiconductor space this eye-popping first-day gain was largely a classic low-float squeeze with only about 7% of shares available to trade driven by heavy state support for chip self-sufficiency and reports that Apple is testing CXMT’s DRAM for local Chinese devices. The company backed up the hype by swinging to a 35.4 billion yuan Q1 profit, benefiting directly from the ongoing global AI infrastructure boom.
while leaders like Micron ($MU) and SK Hynix are shifting cleanroom capacity to high-margin high-bandwidth memory (HBM) for AI, CXMT is aggressively expanding standard DDR5 and mobile memory, targeting 350,000 wafer starts per month by year-end.
However, trade restrictions prevent CXMT from acquiring advanced Extreme Ultraviolet (EUV) tools, leaving its production costs roughly 30% higher than Western peers. Ultimately, chasing low-float, first-day pops near the top of a memory cycle carries severe downside risk. Rather than buying into the hype, retail investors should monitor how CXMT's massive supply expansion impacts global DRAM pricing over the next two years.
China’s DRAM giant, Changxin Technology Group (CXMT), made headlines with a massive 466% surge on its Shanghai STAR Market debut, raising $8.6 billion and briefly becoming China’s most valuable listed stock at a $487 billion market cap. For those watching the semiconductor space this eye-popping first-day gain was largely a classic low-float squeeze with only about 7% of shares available to trade driven by heavy state support for chip self-sufficiency and reports that Apple is testing CXMT’s DRAM for local Chinese devices. The company backed up the hype by swinging to a 35.4 billion yuan Q1 profit, benefiting directly from the ongoing global AI infrastructure boom.
while leaders like Micron ($MU) and SK Hynix are shifting cleanroom capacity to high-margin high-bandwidth memory (HBM) for AI, CXMT is aggressively expanding standard DDR5 and mobile memory, targeting 350,000 wafer starts per month by year-end.
However, trade restrictions prevent CXMT from acquiring advanced Extreme Ultraviolet (EUV) tools, leaving its production costs roughly 30% higher than Western peers. Ultimately, chasing low-float, first-day pops near the top of a memory cycle carries severe downside risk. Rather than buying into the hype, retail investors should monitor how CXMT's massive supply expansion impacts global DRAM pricing over the next two years.
The Kospi's 4.9% plunge - it's the sharpest signal that the AI/chip rally is getting wobbly right before the biggest US tech names report, setting up a high-stakes week.
What are your thoughts on this news?
Workday keeps showing up on "expensive software" screens because of its ~44x trailing P/E. But that trailing number is sitting on a depressed/volatile earnings base, and the forward picture looks nothing like it. Wanted to lay out the bull and bear perspectives.
The bull case:
- Gross margin 75.77% - real pricing power, entrenched in enterprise HR + finance (high switching costs).
- Free cash flow $3.12B, exceeding net income -> high earnings quality, not accounting sugar. That's an ~8.8% FCF yield on a $35B market cap.
- Revenue growth 13.5% YoY, and earnings inflecting hard off a trough.
- Net cash balance sheet (~$548M net) - in a higher-for-longer rate world, that's an income tailwind, not a drag.
- Forward estimates + a low PEG suggest the market is pricing the trough, not the growth.
- Technically in an uptrend (RSI ~62, MACD positive) without being overbought.
The bear case (real risks):
- Trailing P/E of 44x reflects genuine earnings volatility - if guidance slips, the multiple compresses.
- Tight liquidity: current ratio 1.01, quick ratio 0.91.
- Operating margin 13.3% lags software peers (usually 20-25%) - the margin-expansion thesis is a hope, not a fact yet.
Bottom line: BULLISH, medium-term, medium risk - hinging on the earnings reacceleration being real.
Disclosure: All the data and grades above were gathered from ASignal's research tool
CleanSpark is having a good week. Meta is reportedly in talks to lease their Sandersville data center for AI infrastructure, the stock is up and the Bitcoin miner AI pivot story is getting traction. On the surface, it sounds like a business transformation in motion.
But there's a number buried in the financials that most coverage isn't talking about. The gross margin trap CLSK's gross margin is 50.71%. For a Bitcoin miner, that's actually respectable - for every dollar of revenue, 50 cents survive after direct mining costs. If you scan the headlines, this is the number that makes the bull case feel plausible.
The operating margin is -246%
That means for every dollar CLSK brings in, they spend $3.46 all-in. The gross margin is real - the product economics work. But the operating overhead is consuming the business three times over. Revenue could double tomorrow and they'd still be deeply unprofitable...
This distinction
gross margin vs operating margin - is one of the most common ways companies in transition periods look healthier than they are... The product works the business doesn't yet.
What the underlying business looks like right now
- Revenue down 24.9% year over year - the company is shrinking while trading on a growth narrative
- Operating cash flow: -$526M. Free cash flow: -$300M. CLSK cannot fund its own operations
- Total debt $1.79B against $260M cash - net debt of $1.53B in a 5.25% rate environment
- Beta 3.81x - a 10% market selloff corresponds to an expected 38% drop in CLSK
The current ratio is 8.26x, which looks like strong liquidity. It's misleading. Short-term assets don't fix a $526M annual cash burn rate.
Where the frameworks landed
I ran this through three independent investment frameworks - value, activist, and macro - to see if any of them found a different angle, the score and summary is as follows:
Buffett framework (32/100): TOO HARD PILE. No economic moat, negative owner earnings, capital destruction confirmed by ROE of -34.8%. The gross margin doesn't survive contact with operating reality.
Ackman framework (38/100): PASS. FCF yield of -6.8% disqualifies it outright. Ackman's framework requires free cash flow generative businesses. CLSK is the opposite. The Meta talks are unconfirmed and don't change the cash flow structure.
Dalio framework (24/100): AVOID. The combination of 3.81x beta, $1.53B net debt, and -$526M operating cash flow in a high-rate environment creates what Dalio's framework calls a distressed capital structure. This fails every All-Weather criterion.
All three landed independently BEARISH. That kind of cross-framework agreement is rare - usually at least one finds something to work with.
The bull case isn't wrong, it's early
The genuine argument for CLSK is the Meta data center talks. If that deal closes, it means:
- Recurring infrastructure revenue that isn't tied to Bitcoin price
- A path toward operating margin improvement as AI infrastructure scales differently than mining
- Potential rerating as an infrastructure company rather than a miner
That's a real thesis. The problem is "if it closes" is doing a lot of work. Unconfirmed talks don't change the -$526M operating cash flow, don't reduce the $1.79B debt and don't stop the revenue contraction that's already happening. The stock is being priced on what CLSK could become. The numbers reflect what it is right now.
What am I missing?
Has anyone tracked the actual terms being discussed with Meta, or whether similar deals in the data center space have materially changed the unit economics for miners who've made this pivot? Curious whether there's a comparable case where the transition timeline was faster than the fundamentals suggested.
---
(Analysis based on ASignal - I built it. 10+ specialized agents ran this independently, 20+ sources synthesized and digested. Happy to share the full breakdown)
9 weeks in. No co-founder. Building everything alone (I'm a SW engineer with 15 years of experience and with several years in the LLM-AI domain).
The product is called ASignal. It runs a pipeline of specialized AI agents that each evaluate a stock from a different investment framework - one reasons like a value investor, one like an activist, one like a macro trader. Then a separate challenger agent actively tries to poke holes in whatever they agreed on. The output reflects
genuine disagreement between frameworks, not one model's opinion dressed up as insight. I've been posting on Reddit - educational breakdowns of real stocks using the analysis output. Building in public, trying to add value, see what lands.
The feedback I keep getting: "it's just ChatGPT."
Not "the analysis is wrong"... Not "I don't find this useful" just dismissal of the entire thing as one undifferentiated blob of AI hype.
I understand where it comes from. The space has been flooded with wrappers. People have been burned and the skepticism is earned.
But there's a specific kind of frustration that builds when you've spent months on something genuinely complex - multi-agent state passing, adversarial review loops, structured output synthesis across models - and the response is "cool, another AI chatbot."
I'm not looking for validation. I'm trying to understand how other AI founders navigate this. How do you get people to see what you actually built when the category has been poisoned by noise?
Anyone else building in AI solo right now - how are you handling the "just ChatGPT" problem?
This past week, I tried something in this sub: breaking down real companies in plain terms for people learning how to read the market.
One post was on Visa - explaining what a "moat" actually means using their real numbers (97 cents of every revenue dollar is operating margin -that's the moat). One was on SoFi - looking at why "101% earnings growth" and "the stock dropped anyway" can both be true at the same time. Both got almost no engagement.
Before I stop or keep going, I want to ask the community directly: is this type of content actually useful to you, or is it just noise in the feed?
The format I had in mind:
- One company that's in the news right now
- One concept explained without jargon: P/E ratios, operating cash flow, what debt means differently for banks vs tech companies
- What the actual numbers say - not a recommendation, just how to read what you're looking at
Maybe this sub already has enough of this. Maybe the specific companies didn't resonate. Maybe the format needs to change entirely.
If you're here to learn how to actually understand financial news rather than just consume it - what would you want a post like this to cover?
Visa is the stock where everyone agrees on the quality and argues about the price. Here's what the numbers actually say.
The moat case
Network effects are the core: more cardholders → more merchants accept Visa → more people want Visa cards. That loop has compounded for 60 years across 200+ countries. Moat durability estimate: 20+ years.
The financials reflect it:
- Gross margin: 97.78%
- Operating margin: 67.35% (sector median: ~30%)
- ROE: 60.35%
- Free cash flow: $20.84B - importantly, this exceeds net income, which is the quality signal. Earnings can be shaped by accounting (cash flow is harder to fake).
The valuation problem
Current P/E: 28.48x. Forward P/E: 22.02x. PEG: 1.44x. FCF yield: 3.35% - below the 5% threshold most serious Buffett disciples want before initiating a position.
One wrinkle the bulls often skip: Visa is a net debtor. $13.91B cash against $23.98B total debt = -$10.06B net position. In a higher-rate environment, that refinancing risk is real, even with $20B+ in annual FCF.
How Buffett's framework actually scores it
"Wonderful company at fair price" - not a bargain, but not a pass. The logic: at 35% earnings growth you're paying 1.44x the growth rate. For a business with this capital efficiency and moat durability, that's within acceptable range.
Ackman's framework goes further - rates it A+ as a concentrated-bet candidate given the #1 market position and pricing power.
Dalio's framework - macro lens is the dissent: net debtor position + elevated multiples = three-quarter weight, not full conviction.
My summary:
This isn't a deep value opportunity. FCF yield at 3.35% won't satisfy margin-of-safety investors. But for those whose framework includes quality compounders at fair prices - Buffett's Berkshire has held both Visa and Mastercard for years. That's a data point. Entry zone per the technical setup: $320-$330. Support at $310, resistance at $355. R:R is 1:1.6.
Question: how do you all handle the margin of safety problem with wide-moat compounders? Strict FCF yield threshold, or does moat durability change the equation?
(Screenshots not allowed unfortunately, but this analysis came from a multi-agent framework I've been building that runs stocks through Buffett/Ackman/Dalio investor models. Visa scored A / A+ / B+ respectively, landing bullish at 84% confidence. Happy to share the full research if anyone's curious.)