What I find underexplored: how this dynamic is reshaping founder fundraising psychology. When public markets snap back this violently, the private market narrative resets almost in real time too, LP confidence, valuation anchors, risk appetite all shift before the quarter even closes.
Been writing about exactly this tension, how macro whiplash filters down into early-stage decision-making. Happy to share if anyone's thinking through the same.
As a small bakery business with web and four stores, the SAAS software we use is becoming increasingly problematic to achieving AI automation.
SAAS advantage is building a UI for the data it stores or processes. Those UIs have been enshitified over the years as the SAAS providers have been adding features to try and get more monetization and differentiation. That strategy is at odds with what I as a small business want to achieve with AI.
I want to create an AI interface to my data in that is stored in the SAAS that is turn to my use case and allows me to have UIs and agents work across the SAAS platforms. I'm already at odds with my inventory management SAAS provider. The answer is two fold:
1. Build my inventory management from the ground up. Then I can have my own data nsematics and direct database access.
2. move to an open source source platform that enables some UI and existing data semantics but allows me full customization with new UIs and agenic solutions.
The SAAS provider wants me to engage their engineering but they never seem to be able to get a meeting together. Likely because they don't see an answer that maintains their business case.
At least from my perch, the SAAS providers only "moat" is the data stored in the SAAS specific database. Do they allow direct access to that data. If they do, they destroy their moat. If they don't, I move.
The SAAS providers certainly have legacy momentum but as far as I can tell, zero forward strategy for maintaining their business case. Especially given how good AI LLMs are at coding.
I predict that by next tax season, TaxAct and Turbotax will have lost at least half of their subscription base if AI LLMs can connect directly to submit taxes for the federal government and states. That's a very simple concept for them to enable. An the LLM has a much better UI than either of the platforms. Also given the experience I had with taxes this year, AI LLMs are better at implementing the rules and discovering additional tax deductions. I believe TaxAct, TurboTax, H&RBlock and tax accountants are creating a disinformation campaign around the potential problems to slow the inevitable.
I'm only an anecdotal case, but from my perch, the SAASpocolips is real. I have the advantage that I have a deep technical background also. Other small business may not. But I still don't see how a small business wouldn't be looking at AI as a way of lower SAAS subscription costs at the same time it increases useability and productivity.
The speed of the recovery is the data point that worries me more than the recovery itself. Eleven trading sessions to erase a 10% drawdown means the market didnt process the Iran risk. It just priced through it.
PTJ's self-reinforcing loop is the part that deserves more attention than the valuation multiple. At 252% market cap to GDP, a 35% correction isnt just a portfolio problem. Its a fiscal problem. Capital gains revenue disappears, the deficit expands, Treasury issuance increases, bond yields rise, and that feeds back into equity valuations. The correction creates the conditions for a deeper correction.
Buffett sitting on $373 billion isnt a prediction. But it is a price signal from someone who has seen 50% drawdowns three times and thinks this 10% dip wasnt even worth opening the chequebook for.
The problem with an 11-day V-recovery is this: by the time confirmation signals appear, the recovery is already 70–80% complete. People then start buying back in—but at prices close to the highs. They ride the drop AND miss the recovery → a double loss.
This isn’t a one-off. V-recoveries are getting faster. Liquidity, passive flows, and algorithms react in minutes instead of days. Pure experience and gut feeling simply don’t scale on this time horizon anymore.
So the question I wanted to answer was: can you systematically capture V-recoveries instead of missing them?
"Detect V-formations, identify the bottom points, and then ride the trend on SPY (1D)."
Cortix turned that into a strategy, tested it (and even challenged parts of it), and optimized it—without any further input from me.
Backtest over 5 years, SPY, 19 trades:
→ +112% vs. Buy & Hold +84%
→ Max drawdown: −5.6%
That’s just crazy—and exactly what I expect from AI when we use this kind of technology: simplifying complex, data-intensive tasks through an intelligent assistant system.
Super interesting edition! I will say though, the "All Ages Have Cut Back on Scrolling" segment lacks any context about how the data was collected. Were these daily scrolling times self-reported? That matters a lot. Which platforms are included in "social media"? And is this content all video/short-form or does this include people reading X & Linkedin posts? Sources cited for this article were also vague and unhelpful ("CWI, Financial Times").
The wrinkle in semis-first-software-second this cycle: the AI app layer doesn't have to wait for the hardware install base to mature — every developer with API access is on the new platform day one. Mobile software waited 5 years because consumers needed to buy phones. Here the gating constraint is different — distribution, trust, and integration into existing workflows, not capacity. That's part of why the public-private spread is so wide: the apps that exist are scaling fast, the public-market mechanism for valuing them just hasn't caught up.
The semis-then-software framing rings true from where I sit. The wrinkle this cycle: the application layer is already scaling faster than the smartphone era ever did but just mostly in private markets.
Public software multiples probably don't reprice until enterprise buyers move AI from "experiment" line items to "productivity" line items. That handoff is the real catalyst to watch and it's a CFO call, not an engineering one.
Great read! Another thing is stock compensation has been an very important part for SaaS companies, share dilution will be unsustainable for smaller ones if they don’t see revenue acceleration.
What I find underexplored: how this dynamic is reshaping founder fundraising psychology. When public markets snap back this violently, the private market narrative resets almost in real time too, LP confidence, valuation anchors, risk appetite all shift before the quarter even closes.
Been writing about exactly this tension, how macro whiplash filters down into early-stage decision-making. Happy to share if anyone's thinking through the same.
As a small bakery business with web and four stores, the SAAS software we use is becoming increasingly problematic to achieving AI automation.
SAAS advantage is building a UI for the data it stores or processes. Those UIs have been enshitified over the years as the SAAS providers have been adding features to try and get more monetization and differentiation. That strategy is at odds with what I as a small business want to achieve with AI.
I want to create an AI interface to my data in that is stored in the SAAS that is turn to my use case and allows me to have UIs and agents work across the SAAS platforms. I'm already at odds with my inventory management SAAS provider. The answer is two fold:
1. Build my inventory management from the ground up. Then I can have my own data nsematics and direct database access.
2. move to an open source source platform that enables some UI and existing data semantics but allows me full customization with new UIs and agenic solutions.
The SAAS provider wants me to engage their engineering but they never seem to be able to get a meeting together. Likely because they don't see an answer that maintains their business case.
At least from my perch, the SAAS providers only "moat" is the data stored in the SAAS specific database. Do they allow direct access to that data. If they do, they destroy their moat. If they don't, I move.
The SAAS providers certainly have legacy momentum but as far as I can tell, zero forward strategy for maintaining their business case. Especially given how good AI LLMs are at coding.
I predict that by next tax season, TaxAct and Turbotax will have lost at least half of their subscription base if AI LLMs can connect directly to submit taxes for the federal government and states. That's a very simple concept for them to enable. An the LLM has a much better UI than either of the platforms. Also given the experience I had with taxes this year, AI LLMs are better at implementing the rules and discovering additional tax deductions. I believe TaxAct, TurboTax, H&RBlock and tax accountants are creating a disinformation campaign around the potential problems to slow the inevitable.
I'm only an anecdotal case, but from my perch, the SAASpocolips is real. I have the advantage that I have a deep technical background also. Other small business may not. But I still don't see how a small business wouldn't be looking at AI as a way of lower SAAS subscription costs at the same time it increases useability and productivity.
The speed of the recovery is the data point that worries me more than the recovery itself. Eleven trading sessions to erase a 10% drawdown means the market didnt process the Iran risk. It just priced through it.
PTJ's self-reinforcing loop is the part that deserves more attention than the valuation multiple. At 252% market cap to GDP, a 35% correction isnt just a portfolio problem. Its a fiscal problem. Capital gains revenue disappears, the deficit expands, Treasury issuance increases, bond yields rise, and that feeds back into equity valuations. The correction creates the conditions for a deeper correction.
Buffett sitting on $373 billion isnt a prediction. But it is a price signal from someone who has seen 50% drawdowns three times and thinks this 10% dip wasnt even worth opening the chequebook for.
The problem with an 11-day V-recovery is this: by the time confirmation signals appear, the recovery is already 70–80% complete. People then start buying back in—but at prices close to the highs. They ride the drop AND miss the recovery → a double loss.
This isn’t a one-off. V-recoveries are getting faster. Liquidity, passive flows, and algorithms react in minutes instead of days. Pure experience and gut feeling simply don’t scale on this time horizon anymore.
So the question I wanted to answer was: can you systematically capture V-recoveries instead of missing them?
In Vestrix.ai (www.vestrix.ai) I described what I was looking for:
"Detect V-formations, identify the bottom points, and then ride the trend on SPY (1D)."
Cortix turned that into a strategy, tested it (and even challenged parts of it), and optimized it—without any further input from me.
Backtest over 5 years, SPY, 19 trades:
→ +112% vs. Buy & Hold +84%
→ Max drawdown: −5.6%
That’s just crazy—and exactly what I expect from AI when we use this kind of technology: simplifying complex, data-intensive tasks through an intelligent assistant system.
Super interesting edition! I will say though, the "All Ages Have Cut Back on Scrolling" segment lacks any context about how the data was collected. Were these daily scrolling times self-reported? That matters a lot. Which platforms are included in "social media"? And is this content all video/short-form or does this include people reading X & Linkedin posts? Sources cited for this article were also vague and unhelpful ("CWI, Financial Times").
I otherwise love this series, keep it up please!!
The wrinkle in semis-first-software-second this cycle: the AI app layer doesn't have to wait for the hardware install base to mature — every developer with API access is on the new platform day one. Mobile software waited 5 years because consumers needed to buy phones. Here the gating constraint is different — distribution, trust, and integration into existing workflows, not capacity. That's part of why the public-private spread is so wide: the apps that exist are scaling fast, the public-market mechanism for valuing them just hasn't caught up.
Feels like the lag between public and private markets is disappearing. Everything reprices faster now whether people want it or not.
The semis-then-software framing rings true from where I sit. The wrinkle this cycle: the application layer is already scaling faster than the smartphone era ever did but just mostly in private markets.
Public software multiples probably don't reprice until enterprise buyers move AI from "experiment" line items to "productivity" line items. That handoff is the real catalyst to watch and it's a CFO call, not an engineering one.
another smart and informative article
Great read! Another thing is stock compensation has been an very important part for SaaS companies, share dilution will be unsustainable for smaller ones if they don’t see revenue acceleration.