IASV Notion Knowledge Sharing

Date: Friday, April 3, 2026 Participants: Matias Meirelles, Earl Rennison, Dax Craig Jr.

Notion AI -- Capabilities & Limitations

Strong for: deal intake automation, pipeline status updates, meeting note storage, "too early" company reminders
Weak for: external web scraping, data enrichment, deep founder/competitive research
Notion agents released late February; free until ~May 3, then credit-based billing begins
- Current plan ~$20/month business account; credits don't roll over
- Estimated cost with credits + business plan: ~$300/month
Security vulnerabilities already documented in Notion AI -- flagged as a concern for VC use
Recommended hybrid: keep Notion for pipeline tracking and basic automation, use Claude Code for deep evaluation

Proposed custom portal at iasv.co -- role-based access for GPs, admins, and LPs
Key modules envisioned:
1. Deal pipeline (current + historical, 940+ deals reviewed across 4-5 sheets)
2. Portfolio view (37 companies, $3.1M deployed across Fund 1 and Fund 2)
3. AI screening -- batch preliminary screen -> deep evaluation -> deal memo output
4. "Wall of Failure" -- deals passed that succeeded, to feed recursive learning
5. LP portal -- lightweight, high-level overview only (see Earl's feedback below)
Single source of truth goal: consolidate Google Drive, Notion, email, WhatsApp, Ed's memory
Estimated build: 4 phases; Phase 1 largely ready to move on quickly

Strong caution against giving LPs too much information or daily access
- Earl's prior experience: full transparency with hedge fund investors led to confusion and second-guessing of decisions
- "LPs don't need to see how the sausage is made -- that's for GPs/admins"
LP portal should be: light on detail, high-level overview, portfolio companies + updates only
Pipeline visibility for LPs = risk of them questioning deal decisions in real time
Data room-style access (one-off basis) is fine; persistent pipeline access is not recommended
Action: Earl and Matias to sync separately to define LP portal specifications

Contrarian risk: If all VC firms use the same AI tools, they converge on the same deals
- Risk: everyone piles into the same companies (e.g., Perplexity cited as possible example)
AI model inconsistency: Same prompt can return conflicting answers within one response or across runs
- Earl tested same company across Gemini, ChatGPT, Claude -- all said "invest" after model updates
Proposed fix: Build in a contrarian perspective / devil's advocate prompt to stress-test decisions
Proprietary edge: Comes from confidential information others don't have -- RAG over internal files outperforms generic web search
Recursive self-improvement: Feed outcomes (passes, successes, failures) back into screening criteria to refine evaluation over time

YC companies now priced at ~$40M pre-revenue at pre-seed
- Challenges VC math: smaller ownership, more dilution over time
- One success per fund no longer sufficient to return the fund -- need multiple wins
IASV strategy: use YC as a market signal, find similar companies in other markets at better valuations (e.g., FlashPass)

Prepare talking points + brief video walkthrough of the IASV OS for Monday's call with Ed
Present operating system concept in first 5 minutes of the call
Continue using Notion for deal tracking in the interim; export as CSV to inform OS build
Set up IASV GitHub for sharing skills/code (coordinate with Mark)
Send Earl and Dax the Lobster Capital article on YC valuations
Explore Notion agents for what's feasible now (mostly pipeline stage movement + reminders)

Back Matias up on the operating system pitch during Monday's call
Sync with Matias separately to define LP portal specifications -- what to include and exclude

Confirm willingness to pay for Notion AI credits beyond the free tier (deadline ~May 3)
Align on "too early" vs. "hard pass" definition and formalize the 6-month revisit protocol
Input on LP portal requirements

Notes by: Matias Meirelles Transcript: Granola