Every position graded against its own stated thesis - not the price alone.
A thesis is broken below its invalidation level, testing within 3% of it, else on track.
Belief state lives in scripts/theses.json - edit it directly, the page re-renders on the half hour.
Largest single-name conviction. Autonomy/robotaxi optionality on top of a volume+energy business; held through volatility by design, derivatives used only as defined-risk overlays.
Invalidation $298 ·
price +23.5% above · Loss of the 6-month floor (~298) alongside shrinking auto gross margin = the volume story breaks; autonomy alone has never carried the multiple.
- ☐ Auto gross margin ex-credits stable or rising
- ☐ FSD/robotaxi milestone cadence continues
- ☐ Energy storage deployments growth intact
Agentic-trading + wealth-transfer thesis: the retail brokerage layer that agents and the next generation of investors flow through. Trimmed Aug 25 from ~32% to still-largest book weight; re-add only if adoption metrics arbitrate the Gerstner-trim vs Camillo-max-conviction divergence in favor of hold.
Invalidation $88 ·
price +10.0% above · Below the pre-summer range top: would require funded-customer growth and trade-per-user both decelerating at once - the two legs the thesis stands on.
- ☐ Funded customers + ARPU accelerating, not just price
- ☐ Agentic/API order flow becoming measurable
- ☐ Net deposits trending up
AWS reacceleration + retail margin expansion + ads; the diversified mega-cap compounding core. External conviction: Camillo's largest-ever position (leveraged, ~70%). Keep full LEAP structure; adds only below the $248 entry ceiling.
Invalidation $240 ·
price +9.7% above · AWS growth rolling over while retail margins compress - the twin-engine thesis fails and the stock becomes a multiple story, not an earnings story.
- ☐ AWS growth >= prior quarter
- ☐ Retail operating margin expansion
- ☐ Capex translating into backlog, not just spend
Electrons-not-algorithms: contracted AI infrastructure capacity. Trimmed Aug 25, not exited - the neocloud leg stays on while contracts, not narratives, prove out.
Invalidation $165 ·
price +4.4% above · High-beta name (6m range 135-286): below 165 with contracted-capacity growth stalling, the story becomes dilution-funded capex with no operating leverage.
- ☐ Contracted capacity (MW) guidance raised, not cut
- ☐ Customer concentration improving
- ☐ Funding roadmap without heavy dilution
The AI capex king and pricing power anchor of the whole book's theme. Position deliberately small vs TSLA/HOOD - it is the ballast, not the bet.
Invalidation $190 ·
price +14.5% above · 6-month floor (~189): a break means hyperscaler capex guidance has plateaued - the one macro input the entire AI complex prices off.
- ☐ Hyperscaler capex guides rising, not flat
- ☐ Gross margin holding through ramp
- ☐ Top hyperscalers stay >50% of revenue
Custom AI silicon #2: hyperscaler ASIC second-source. The concentrated book's diversification into the one capex stream guaranteed to grow.
Invalidation $180 ·
price +32.6% above · Below 180 with hyperscaler design wins slipping to 2028: the ASIC ramp is later than priced and the stock is a story, not numbers.
- ☐ New socket wins announced each quarter
- ☐ Data-center revenue growth > 20% y/y
Gemini+TPU vertical integration - the only player owning the whole AI stack from silicon to distribution. Position token-sized; thesis optionality, not size.
Invalidation $300 ·
price +15.0% above · Below 300 with search-query share eroding to AI assistants: the distribution moat is wrong and TPU external wins price as one-offs.
- ☐ Search query volume growing through AI transition
- ☐ TPU external customers signed
- ☐ Gemini usage closing the gap
Leveraged wrapper on the TSLA core - convexity, not a separate thesis. Lives and dies with TSLA above; the only independent variable is the decay/vol drag budget.
- ☐ Gate 2026-10-24: net-long delta reduction executed
- ☐ Decay drag vs TSLA move ratio acceptable
Exited Aug 25 (profit-taking batch). Re-entry is optionality, not conviction: only if the Ampere-commitment / NVIDIA-financing bull points prove out in signed contracts.
Re-open trigger: Signed Ampere/NVIDIA-financing contract evidence; wash-sale window (to 2026-09-23) has cleared.
- ☐ Contract announcements, not LOIs
- ☐ Debt cost vs contracted-revenue spread
Small energy satellite from the Aug 25 batch (trimmed, held). No independent thesis on record yet - treat as harvest candidate on strength.
- ☐ Write the thesis or harvest it - limbo is the only wrong state
Nano-position (~800 sh @ ~$6). No thesis on record - candidates: write one, average in on evidence, or exit on the next portfolio clean-up.