Marcus T

Very questionable ethics imo, so while it may be a moonshot in the short run, I am not confident that this will last the test of time, especially when copycats start popping up, or if big pharm start taking actions. But we’ll see

Our CIO wrote this over the weekend on how agentic AI is good thing for software platforms. A good casual read... Especially if you are in software land. $CRM $TEAM $NOW $IGV

$BTC get out while you can. I rather have $GLD, which is 15% of my portfolio

Counterpoints: 1. Platform differentiation trends toward zero Rebuttal: While AI makes it easier for a startup to write code, it does not give them the 15 years of proprietary customer data, complex permission structures, and custom workflows locked inside a Salesforce or ServiceNow. If everyone has the same LLM, the winner is the platform that feeds the LLM the best, most structured proprietary data. Incumbents own the "Context Window" of the enterprise. A startup can replicate the interface, but they cannot replicate the data graph required to make the AI accurate. 2. Value accrues to the agentic layer sitting on top Rebuttal: The System of Record (SOR) *is* the Agentic Layer. The idea of a thin "agentic layer" floating above the System of Record ignores enterprise security and latency. CIOs will not authorize a third-party agent to read/write into their critical financial or HR databases without strict governance. Instead, the SORs are building their own agentic layers (e.g., Salesforce Agentforce). The value accrues to the incumbent because they own the security perimeter required for the agent to actually execute tasks. 3. Investor sentiment becomes a structural headwind Rebuttal: Sentiment follows FCF growth, and AI is the ultimate efficiency lever. The current de-rating is based on uncertainty. However, top-tier SaaS companies are using AI to flatten their own org charts—automating support and entry-level engineering—which (should) drastically reduce their opex. We are entering a "Golden Age" of margins for incumbents who use AI to run leaner while charging customers more for AI features. Once the "AI revenue" line item proves to be margin-accretive, the multiple expansion will return. 4. AI-native startups eat incremental LTV with better prices Rebuttal: The "Bundle" crushes the "Point Solution." Startups have a distribution problem; Incumbents have a feature problem. It is infinitely easier for Atlassian to add "AI coding agents" to Jira than it is for a startup to build an enterprise sales force, get SOC2 compliance, and displace a trusted vendor. CIOs will simply pay a modest upsell to a trusted vendor (zero marginal CAC for the incumbent) rather than onboard a risky new vendor for a "cheaper" price. 5. Seat-based revenue will decline Rebuttal: History shows that when you reduce the friction of a task, demand for that task increases. As AI makes employees more productive, companies often hire more people to attack higher-order problems. More seats! Furthermore, top-tier SaaS is already pivoting to consumption/outcome-based pricing (per conversation, per resolution). This allows them to capture the value of the digital labor directly, likely expanding the TAM beyond what human "seats" could ever cap out at. 6. Legacy SaaS will struggle to transition to outcomes Rebuttal: They are the only ones capable of selling outcomes. To sell an "outcome," you must be able to measure the "before" and "after." The System of Record holds the historical truth. ServiceNow knows exactly how long an IT ticket took to resolve in 2023 vs 2026. They are in the absolute best position to prove value and charge for the outcome because they own the ledger that defines what an "outcome" is. A startup has no baseline data to prove they delivered the result. 7. Diminished pricing power and lock-in Rebuttal: AI creates "Intelligence Lock-In." An AI model is generic, but an AI model trained on your specific 5-year history of Jira tickets or Sales Cloud interactions is a proprietary asset. The longer you stay with the incumbent, the smarter the AI gets about your specific business. Moving to a competitor doesn't just mean migrating data rows anymore; it means lobotomizing your organization's institutional memory and retraining a brain from scratch. 8. Gross margins will deteriorate Rebuttal: This assumes the cost of compute stays static. Instead, inference costs are dropping exponentially. Meanwhile, top-tier SaaS companies have ~80% gross margins and can absorb the initial opex to secure the market. Unlike startups, they can negotiate wholesale compute deals with hyperscalers. Over time, they will charge a premium for AI features (e.g., $30/user/month) that vastly outstrips the plummeting cost of the inference required to run them. 9. Decreased organic traffic increases CAC Rebuttal: Irrelevant for Enterprise SaaS. This point is a valid fear for PLG (Product-Led Growth) tools or SMB software, but Workday and ServiceNow do not close 7-figure enterprise deals because a CTO googled "HR software" and clicked an SEO link. They close via field sales, partner channels (Accenture/Deloitte), and multi-year relationships. LLMs answering search queries changes nothing about the high-touch enterprise sales motion. 10. Competition for talent increases SBC/Opex Rebuttal: Application AI is not Research AI. The "talent war" is for the 0.001% of researchers building Foundation Models (OpenAI, Anthropic). SaaS companies just need engineers who know how to use APIs and orchestrate models—skills that are becoming standard. Furthermore, AI makes the incumbent’s existing engineering team 30-50% more productive. They can ship more product with less headcount growth, which neutralizes the cost of talent.

Bought 3x LEAP $QQQ puts $625 today. We’re probably very near the top.

Bought so much of $GOOG at average price of $116 that it’s now my 2nd top individual stock. Not going to buy more for now.