Skip to content
Tuesday, September 1, 2026
VideosArchive
Gross Retention
The Book

A portfolio of published theses, traded automatically.

A concentrated enterprise software long/short book. Every position is a published thesis. Positions open and close automatically based on conviction, signals, and market data. Not investment advice.

Benchmark

iShares Expanded Tech-Software ETF (IGV) — shown for context, not attribution. The book targets absolute returns; IGV is the closest investable enterprise software proxy.

Cumulative since Aug 27, 2026. Not annualized.

Portfolio (cum.)+1.56%
IGV+6.39%
Alpha-483 bps
DatePortfolio (cum.)IGVDelta
Aug 270.00%0.00%0.00pp
Aug 28+0.90%-0.89%+1.79pp
Aug 29+1.63%-1.47%+3.10pp
Sep 1+1.56%-0.95%+2.51pp

Exposure

LegUSDActualTargetDrift
Gross$73,95973.3%80%▼ 6.7pp
Net-$6,261-6.2%0%▼ 6.2pp
Overlay long$33,84933.6%40%▼ 6.4pp
Overlay short$40,11039.8%40%▼ 0.2pp
Total P&L (cum.)$1,072 (+1.1%)
Equity$100,866
Gross MV$73,959

Positions

PositionThesisValueUnreal.Held
ADBElongAdobe's Firefly generative AI is not a threat-replacement narrative — it is a moat-deepener. By training exclusively on licensed Adobe Stock content and embedding generation directly into Photoshop, Premiere, and Illustrator, Adobe turns generative AI into a workflow-lock rather than a disruption vector. Enterprise adoption of Firefly Enterprise and Express positions Adobe to capture GenAI content creation budgets.Wrong if: Adobe Digital Media ARR growth falls below 8% YoY for two consecutive quarters through FY2027$4,955▲ +4.9%—/120
CRMlongSalesforce's Agentforce platform represents the most credible enterprise AI agent monetization story in software. Unlike competitors selling API credits, Salesforce embeds autonomous agents into its existing CRM workflows — sales, service, marketing — and prices them per conversation. This usage-based AI overlay on top of a subscription core gives Salesforce a compounding revenue model that consensus has not fully modeled.Wrong if: Salesforce Data Cloud and Agentforce combined fail to reach disclosed revenue targets by Q4 FY2027$5,954▲ +21.7%—/120
CRWDlongCrowdStrike's Falcon platform survived the July 2024 outage because switching endpoint security is operationally prohibitive — customers must test policies, replace agents, retrain teams, and accept migration gaps that can expose them for months. Subscription revenue remained 95% of total at $1.32B quarterly and ARR reached $5.51B. Platform expansion into identity protection, cloud security, and next-gen SIEM deepens retention.Wrong if: CrowdStrike ARR growth falls below 25% or net new ARR turns negative in any quarter through Q2 FY2028$3,291▲ +15.8%—/120
DDOGshortDatadog's usage-based pricing model — a tailwind when engineering budgets were expanding — becomes a structural headwind as AI agents generate telemetry volumes that explode beyond linear growth but without proportional budget expansion. Customers facing unpredictable bills will cap, sample, or migrate observability workloads to cheaper alternatives.Wrong if: Datadog revenue growth stays above 25% AND net retention stays above 120% for four consecutive quarters through Q2 FY2028-$9,892▼ +0.6%—/120
GTLBshortAI coding assistants that generate complete applications from natural language prompts bypass GitLab's DevSecOps platform entirely. When a product manager can prompt an AI to build, test, and deploy a full application without writing code, committing to Git, or configuring CI/CD pipelines, the entire DevOps toolchain value proposition faces obsolescence risk.Wrong if: GitLab revenue growth stays above 25% and dollar-based net retention stays above 120% through Q2 FY2028-$5,473▼ +9.4%—/120
NETlongCloudflare's edge infrastructure positions it as the inference and security layer for the distributed enterprise — running models and enforcing policy at the edge, close to users, on infrastructure already deployed across 330+ cities. Its existing global network becomes a competitive asset that hyperscalers cannot replicate at comparable unit economics without building comparable edge footprints.Wrong if: Cloudflare revenue growth falls below 25% or dollar-based net retention drops below 115% in any quarter through FY2027$2,678▲ +4.8%—/120
NOWlongServiceNow's moat is not AI — it is the installed workflow graph: years of accumulated permissions, dependencies, approval chains, and institutional scar tissue that makes its platform the control plane for enterprise operations. As enterprises deploy AI agents, the unglamorous plumbing of identity, policy, system state and audit trails becomes the critical asset. Subscription revenue is 96.8% of total, cRPO grew 24.5% YoY, and the Now Assist $1B ACV target is on track.Wrong if: ServiceNow subscription revenue growth falls below 20% or cRPO growth drops below 20% in any reported quarter through FY2027$7,093▲ +12.3%—/120
PATHshortUiPath's RPA platform is structurally vulnerable to AI agent displacement. The RPA value proposition — automating repetitive GUI-based tasks through screen scraping and workflow rules — is the exact category most exposed to LLM-based automation that understands context rather than mimicking clicks.Wrong if: UiPath ARR growth reaccelerates above 20% and net retention rises above 115% for two consecutive quarters through FY2027-$5,581▼ +11.9%—/120
PLTRshortPalantir's government-to-commercial pivot faces a growth reality check as large enterprise deals slow and AI-platform competition intensifies from Microsoft, Databricks, and Snowflake. The premium multiple prices in flawless execution that the next two quarters are unlikely to deliver.Wrong if: Palantir revenue growth stays above 25% and commercial customer count grows 20%+ through FY2027-$9,443▼ +6.3%—/120
SNOWshortSnowflake's proprietary data format moat is eroding as Apache Iceberg becomes the industry-standard open table format. When customers can store data in open Iceberg tables accessible by any compute engine, Snowflake loses the storage lock-in that justified its premium pricing. If data is portable, compute pricing must compete on commodity terms, compressing Snowflake's gross margins and growth rate simultaneously.Wrong if: Snowflake product revenue growth reaccelerates above 30% AND net revenue retention stays above 125% through Q2 FY2028-$9,721▲ +0.0%—/120
WDAYlongWorkday's HR and finance data model gives it an architectural advantage in enterprise AI adoption that competitors stitching together applications cannot match. With 92.6% subscription revenue, the Insperity partnership unlocks mid-market distribution while agents that touch payroll, benefits, and financial approvals require governed context only Workday's unified platform provides.Wrong if: Workday subscription revenue growth falls below 15% in any quarter through FY2027$4,811▼ +0.2%—/120
ZSlongZscaler's zero-trust architecture — cloud-delivered policy enforcement between users, applications and the internet — becomes more valuable as enterprise perimeters dissolve. Deferred revenue grew 32% YoY, six percentage points faster than recognized revenue at 26%, signaling contract duration expansion the income statement has not yet captured. The architectural moat from replacing appliance estates creates switching costs that outlast competitive feature parity.Wrong if: Zscaler calculated billings growth falls below 20% in any quarter through FY2027$5,067▲ +2.3%—/120

Costs

Synthetic accrual

Borrow $0.00 · Margin interest $0.00 · Total $0.00

Synthetic. Paper accounts model neither borrow nor margin interest; both are accrued in our own ledger. Shorts are gated to shortable and easy-to-borrow names rather than modelled.

Methodology

Structure. Enterprise software long/short — 7 conviction-weighted long positions at 4,000 bps total exposure, 5 short positions at 4,000 bps. Every position is a deliberate bet on enterprise software — CRM, workflow automation, creative tools, cybersecurity, HR/finance, edge infrastructure — backed by a published thesis and primary research. No consumer, no semiconductors, no e-commerce. Cash funds the short side. Gross exposure 8,000 bps (4,000 long + 4,000 short against 100% equity).

Timing. A position opens at the next regular-session open after publication, never before. Orders are market-on-open, so fill prices are determined by the market, not by when the bot happens to run.

Concentration. Long positions range from 500 to 650 bps depending on conviction. Short positions are capped at 800 bps per name. The book is deliberately lopsided — 7 longs vs. 5 shorts — because strong short ideas are scarcer and sizing should reflect that scarcity rather than dilute conviction.

Shorts are gated, not modelled. Borrow cost is unobservable to us, so the short universe is restricted to names the broker reports shortable and easy-to-borrow, with a conservative flat accrual.

Reporting. Returns are absolute, not benchmark-relative. The book holds no index, so every basis point of return is attributable to the positions on the page. Costs are synthetic — paper accounts model neither borrow nor margin interest — and are accrued in our own ledger.

This portfolio is simulated. It trades a paper account. Nothing here is investment advice.

Broker state as of 2026-09-01 14:11 UTC

Activity Log

Loading activity log\u2026