All Memos
7 example stress-tests across stocks, funds, and private companies. Click any to explore the multi-agent output, or sign in to run your own.
Examples
Read-only · ships with the app- PrivateSeries B·Helix Compute
Helix Compute is GPU virtualization and inference orchestration for mid-market enterprises ($50M–$500M revenue) deploying internal AI workloads. The wedge: fractional GPU allocation across a customer's existing capacity, reducing inference cost 40–60% vs. dedicated tier. Currently $18M ARR up from $5M ($3.6x YoY), 145% NRR, three Fortune-1000 lighthouse customers. Founders ex-NVIDIA (Diana Foster, ex-DGX product) and ex-Run:AI (Krish Iyer, ex-engineering). Series B led by tier-1 infra fund at $200M post.
Changes requested - PrivateSeries A·Lattice Forge
Lattice Forge is an AI-native CI/CD platform for engineering teams of 50–500 engineers. The wedge: failed builds and flaky tests are routed to a specialized fine-tuned model that proposes root-cause patches in <60 seconds. Currently $3.2M ARR up from $400K twelve months ago (~8x YoY), 92% gross retention, 138% net retention. Founders ex-CircleCI and ex-GitHub Actions. Series A led by a tier-1 infra fund at $40M post.
Approved - PrivateSeed·Modular Care
Modular Care is an API-first platform connecting independent primary-care clinics with specialist networks. The wedge is utilization analytics for Medicaid managed-care contracts — a segment where incumbents struggle with low-margin operational reporting. ARR is ~$180K with three regional MCO contracts and 12 clinic-network LOIs. Stage-appropriate team (ex-Oscar, ex-Aledade) with deep domain. $50K check at $8M post-money.
In review - FundMF · Parag Parikh Flexi Cap Fund
Parag Parikh Flexi Cap should compound at 13–16% INR over the next 3–5 years with materially lower max drawdown than the Nifty 500 TRI, driven by three structural edges that domestic-only flexi-caps can't replicate. First, the Indian core (HDFC Bank, Bajaj Holdings, Power Grid, Coal India, ITC — ~34% of the portfolio) is value-tilted on FCF yield and skews to cash-generative incumbents, which historically holds up in rate-cut and earnings-deceleration regimes. Second, the ~15% global sleeve in mega-cap US compounders (MSFT, GOOGL, META, AMZN) gives unitholders rupee-denominated participation in the AI capex / advertising / cloud cycle that Indian indices structurally lack. Third, the team — Rajeev Thakkar, Raunak Onkar, Raj Mehta — has been in the seat for a decade with high skin-in-the-game, and the fund's willingness to carry cash means the dry powder is real, not rhetorical, at corrections.
Rejected - StockSNDKSandisk Corporation
SanDisk has transformed from a cyclical NAND vendor into the cleanest pure-play on AI infrastructure storage, and the fiscal Q3 print released April 30 makes the thesis tangible. Revenue of $5.95B grew 251% year over year, gross margin expanded to roughly 78%, net income swung to $3.6B, and EPS of $23.41 beat the $14.66 consensus by nearly 60%. Datacenter revenue grew 233%, the company is sitting on around $42B of long-term customer commitments, and management authorized a $6B buyback. Industry reports indicate 2026 NAND supply is effectively sold out and that contract prices are stepping up 70 to 75% quarter over quarter, which gives SanDisk pricing power it has not had in any previous cycle. Post-spinoff from Western Digital, the balance sheet is healthy with about $1.48B in cash against $2.0B in long-term debt, and Nasdaq-100 inclusion has structurally widened the buyer base. The setup looks more durable than a normal memory cycle because the demand driver is hyperscaler training and inference storage rather than consumer SSDs, and every incremental GPU cluster needs orders of magnitude more high-density flash. SanDisk's BiCS8 roadmap and Kioxia joint venture give it manufacturing scale that smaller players cannot replicate near term. Sell-side has caught up but is not capitulated: Goldman raised its target to $1,200, Bernstein's bull case sits at $1,700, and Forbes/Trefis has framed it as the best AI play of 2026 on the data-exhaust thesis. The risks are real, with the stock up roughly 470% year to date, genuine pull-forward risk in customer pre-pays, and a history of NAND overshooting in both directions. But with the company printing $23 quarterly EPS, multi-year contracts providing visibility, and a $6B buyback putting a floor under the float, the risk-reward at current levels still skews to the upside for investors willing to hold through volatility.
Changes requested - StockASMLASML Holding N.V.
ASML is going to the moon because it is structurally significant to chip manufacturing and has no competitors
Approved - StockMRVLCurrent price of MRVL
This should compound 18% over 3 years on premium-mix shift, and rural recovery.
Approved